If I ran a retail media network

Every network in the category reports a sales number that no independent party has ever checked. The first one to put that number through an audit owns the only unoccupied position left.

You know the quarterly review. Six retail media networks on six slides, each with a return-on-ad-spend figure next to it, and somebody in the room asks whether 4.1 at one retailer is better than 3.7 at another. The honest answer is that nobody in the room can tell, including the networks, because the two numbers were calculated by different parties using different windows against different definitions of a sale. The meeting moves on anyway, because the alternative is admitting that the largest growing line in the budget is allocated on figures that cannot be compared.

I spent last week reading what has actually been independently checked in this category. The answer turned out to be narrower than I expected, and it points at a move no network has taken.

The audit stops before the number that matters

Amazon's accreditation covers Sponsored Products clicks and DSP impressions, clicks and viewability. Walmart Connect's covers Sponsored Search impressions, clicks and CTR, granted September 2025. Criteo's covers onsite display rendered impressions and clicks, from March 2024. Instacart's covers impressions, clicks, CTR and viewable impressions, continued in June 2026. Roundel and Albertsons Media Collective are not in the MRC's digital directory at all.

Every one of those letters certifies that an ad was served. None of them touches the sales figure the budget was approved against. An advertiser reading an accreditation badge on a retail media deck is reading a guarantee about delivery and inferring one about outcomes, and the gap between those two things is where the entire category sits.

One network has moved. Kroger submitted attributed sales, incremental sales, aROAS and iROAS for MRC accreditation, and has been in pre-audit since June 2026. Nothing has been granted. It is the first outcome metric in the council's retail media pipeline, and it came from a network with roughly a tenth of Amazon's share rather than from either of the two companies taking most of the money.

Read that against what incrementality is sold as. It is positioned as the rigorous alternative to attribution, the number that separates demand a brand created from demand it merely caught. Unaudited, it is the opposite: a figure with more discretion inside it than ROAS, because it has more construction inside it. Three private methodology choices sit between the raw data and the headline, and all three belong to the party being graded.

Audited outcomes, and the data to check them

So here is what I would do with a challenger network, and it is one move rather than three. Submit the outcome metrics to independent audit, report incrementality as the default rather than as a premium tier, and let the advertiser pull log-level data out and re-derive the lift in their own environment. Those three only work together. Audited methodology without portable data is a certificate the buyer still cannot check. Portable data without an audit is a pile of logs and an argument. Incrementality without either is what the category already has.

The objection arrives immediately, and it is real. Audited numbers are smaller numbers. A network that reports honestly competes against six that do not, and the first quarters are worse. I went looking for evidence on how that trade actually lands and found none — no retailer in this set discloses retail media revenue separately enough to show what a transparency move did to it. That cost is unmeasured, and anyone telling you the move is safe is guessing.

What is measurable is the demand. Advertisers name performance measurement, modeling recommendations and business plan adherence as their top three budget-allocation factors, all at 38%. Performance already moves the money. It is simply that no verified performance exists in this category for it to move toward.

The part that is not about honesty

Transparency arguments lose to performance arguments, and they deserve to. A buyer holding a number that looks good does not want a lecture about where it came from. If this were only an ethics pitch I would not run it.

But a network cannot improve against a measure it writes. Optimization needs an objective the optimizer did not author. A network tuning toward a figure whose methodology it controls, and which moves 6.5x on those controls, cannot distinguish getting better from counting differently — and across enough cycles the system learns the counting. That is a capability ceiling, and it applies to networks acting in perfectly good faith.

Which turns the sales conversation around. Against a closed-loop competitor reporting a higher return, the line is not our numbers beat theirs. It is that theirs cannot be checked, ours can, and only one of us has a mechanism for getting better at this. That argument survives losing the number comparison, because the number is the thing in dispute.

The networks doing the most are not winning

Honesty about the state of play: the measurement leaders in this category are the challengers, and the challengers are losing. Amazon took 79.7% of US retail media spend in 2025 and Walmart 8.0%. Of roughly $10.53bn in net-new US spend forecast for 2026, $9.42bn — over 89% — goes to those two, with every other tracked network expected to hold flat or shrink through 2027. Neither has an accredited sales metric between them.

That is the strongest case against everything above, and the data currently wins it. What the data cannot say is whether measurement would move share, because nobody has shipped the thing being argued for. Instacart has accredited delivery metrics and an open clean room, and no accredited outcome. Albertsons has the best-designed incrementality in the set and no accreditation at all. Kroger has the outcome metrics in audit and stands alone there. Each has a piece. A null result against a treatment nobody has administered is a null on the market, not on the idea.

What NBC actually bought on election night

NBC took 41.5% of its primetime audience from the bracket advertisers pay for, nine points clear of the field. The cheapest thing on the air is getting the credit, and nobody has checked whether it earned it.

Budgets for 2028 do not get written in 2028. They get framed in the weeks after a result, while the result still feels like it means something, and the framing outlives the memory of the night that produced it. That is happening in network news buildings right now.

On November 3, NBC led the primetime field with 6.74 million viewers. It was the network’s first broadcast win on a federal election night in total viewers since 2012, and a jump from fourth place in the 2022 midterms. Across the trade coverage, the thing being credited is Steve Kornacki’s whiteboard and the ten-and-a-half-hour livestream built around it.

That credit is worth something only if it can be told apart from three other explanations that fit the same numbers equally well. Nobody has done that work, and the budget conversation is not waiting for it.

The win is bigger than the headline, in the measure that matters

Nobody sells total viewers. Advertising is sold against a demographic, and the one network news trades on is adults 25 to 54. Divide each network’s demo figure by its total and the night looks different from the way it was reported.

The split is not broadcast against cable, which is the explanation that would make this unremarkable. CNN sits above ABC on the same measure. Fox News and MS NOW sit together at the bottom, seventeen points below a broadcast network that finished last in total viewers.

Four of the six finish in a different position depending on which column you read. Fox News is third in total viewers and fifth in the demo. CBS travels the other way, from last to fourth. For a sales team that is not a technicality — it is the gap between what a network announces and what it can charge for.

One caveat belongs here rather than at the bottom. That percentage is arithmetic on two published figures, not a Nielsen currency. Total viewers counts age 2+, so the ratio is not a measured composition and no Nielsen product reports it. It is the closest published proxy for what a sales team prices, and it is not that thing.

The field moved underneath all of it

Against the 2022 midterms, Fox News fell 42% in total viewers and 59% among adults 25 to 54. CNN rose 59% and 43%. Total primetime coverage across 13 networks reached 31.2 million, up 23% from 25.43 million.

So a network can post its best election night in fourteen years during a cycle when the whole field grew by nearly a quarter and its largest competitor lost more than half of its paying audience. Every one of those facts is true at once, and only one of them is about Steve Kornacki.

That last comparison carries a problem of its own. Big Data + Panel became the currency that settles national television transactions on September 1, 2025, and it runs higher in aggregate than the panel-only basis it replaced. No report of the 23% rise says whether the 2022 figure was restated onto it. Some unknown share of that movement may be the ruler rather than the audience.

Where those four hold, the win is unattributed and the budget decision rests on a story. Where they do not — where the tactic is inseparable, the field was flat, or you have already run the comparison — this pattern is not yours and the credit can stand.

Three networks won the same night

All three cable networks published a most-watched claim from the same Tuesday, and the order changed twice in three days. None of those numbers is the one either side of a deal actually signs.

Three cable networks put out a release this week saying they won election night. All three are correct. That is not a dispute about facts, a correction waiting to happen, or anybody shading a number — it is three different questions, each answered accurately, each published as though it were the question.

What makes this more than a press-release curiosity is the second thing that happened. Between Wednesday and Thursday, the order changed twice, and the number that finally settled it is the same basis that decides whether a network owes its advertisers make-goods in March.

Each network led a different cut of the same Tuesday

Fox News is first on one measure and fifth of six on another. MS NOW leads the eight-hour window and finishes last in the demo. Each release named the superlative and not the cut it rested on.

The full-window figures have two published versions — a second report of the same 6pm–2am window gives 3.3, 3.0 and 2.8 million for the same three networks. That disagreement is still open, and resolving it by picking the more convenient one is how a number becomes folklore.

Nobody’s audience changed between Wednesday and Thursday

Wednesday’s Fast National had CNN first at 3.927 million, MS NOW second at 3.922 million, Fox News third at 3.882 million. A preliminary update later put MS NOW on top. Thursday’s final Big Data + Panel put Fox News first at 4.315 million.

Three releases, three orders, one night. These are not corrections of each other: the Fast National is an early read issued the morning after, and the final combines Nielsen’s panel with large-scale set-top and smart-TV data. Both are Nielsen numbers. They are not the same measurement, and when 45,000 viewers cover three networks, the difference between methods decides the ranking.

None of those figures is what settles a deal

National television inventory is overwhelmingly sold in the upfront, months ahead, against demographic guarantees rather than named programs. The unit has been the average commercial minute since 2007 — C3, the audience for the average minute of advertising inside a program, live plus three days of playback, typically adults 18 to 49. C7 extends that window to seven days.

Election night was reported in program ratings, in total viewers and adults 25 to 54, across a named three-hour block. Different metric, different demographic, different window. Three networks spent three days arguing about a scoreboard that is not the thing either side signs.

And the floor underneath it moved in August. Big Data + Panel became the official currency for national transactions on September 1, 2025, and upfront guarantees are built on it; in aggregate it produces higher impressions, which means fewer audience deficiency units owed to advertisers — though for some networks and dayparts it produces lower ones and more make-goods. Then on August 31, 2026, nine weeks before this election, Nielsen deployed seven simultaneous methodology changes to that service, tied to retaining its accreditation, with no assurance that ratings would rise.

Where those hold, the press number and the commercial number have come apart and the public argument is being fought on the wrong one. Where there is a single agreed measure that both the trade and the contract use, this does not apply and the scoreboard is the scoreboard.

The Playbook Didn't Fail. The World It Described Did.

Classical brand management makes a business legible, and legibility gets mistaken for control. Its own accounting is what makes marketing look free to cut.

This week I could not finish a problem set. A plan bridge for a hypothetical launch: cannibalization, incremental volume, a contribution line falling while sales grew. Three steps in, I stopped, and had to write down the sentence the rooms I have sat in reliably punish — I don't know.

Then I learned it. I spent five years at General Mills in brand experience strategy, working alongside the general managers who owned those P&Ls. I influenced the P&L for five years. I never owned one. So the homework is late and deliberate, and what it turns up is that the two most-complained-about failures in modern marketing are not failures of character. They are outputs of the accounting.

The system was built for a specific physics

Those assumptions were so stable they became invisible. Every one of them has since expired.

Legibility becomes overconfidence

The system's great strength is that it makes a business legible. Everything becomes a bridge, a waterfall, an elasticity, a forecast. Its great weakness is that legibility gets mistaken for control.

A plan that adds up feels like knowledge about the future. It is knowledge about your assumptions. Forecasts and elasticities are estimates with wide error bars, but the dialect of the brand review states them as points of fact, and the room rewards the crispest decomposition rather than the most honest account of what is unknowable.

Add incumbency. Generations of classical managers were trained in categories where the biggest brands won for decades for structural reasons — distribution, scale, media weight. The system credited the operator. Fluency in the language became indistinguishable from being right. In my agency and strategy years I sat in rooms where asking how do we actually know that? made me the slow one. Nobody could show me the count.

The accounting eats the future

Underinvestment in marketing is not a mystery inside this system. It is arithmetic.

When a year goes off plan, something has to give. Trade spend cannot be cut without risking distribution. COGS is factories and commodities. The one large, discretionary, in-year lever is A&P. So the standard rescue is to cut demand creation and deliver the contribution target. Cut $8M of A&P, and a brand running $7M behind plan lands $1M ahead — having improved nothing. Volume down, sales down, margin rate down, and the business "saved" on the single line the cut touches.

Why is this so easy to do? Because brand equity is an asset that lives off the P&L. Harvest it and no depreciation shows. The system underinvests in marketing because its own accounting makes marketing look free to cut.

Innovation gets the same treatment in reverse. A new initiative is charged in full: launch support, distribution costs, the sales it cannibalizes from your own products, the contribution it dilutes in year one. The base business is never charged for standing still while the channel changes underneath it. A digital capability whose return is learning, audience or presence on a new shelf will fail a hurdle designed for a line extension — not because the return isn't real, but because it arrives in a currency the model doesn't count.

What the classical metrics cannot see, measured

A brand can own the answer to "what did we announce." It cannot own the answer to "should you choose us." A content budget cannot move an answer that was never retrieved.

Twelve preference questions across six categories is one observation each, on one assistant, on one day, and the measured retrieval payload doubled in the two days before that run. That locates a boundary. It does not size it, which is why the figure above is a count and not a percentage.

The demand physics did not change

Here is where the "classical is dead" argument overreaches. Penetration still beats loyalty as the growth lever. Mental and physical availability — being easy to think of and easy to buy — still drives choice; the Ehrenberg-Bass work on this has survived every channel shift thrown at it.

And contribution discipline turned out to be the thing a decade of growth-at-any-cost digital brands died for lack of. They bought revenue, ignored payback, and rediscovered at scale what any classically trained brand manager could have told them from a single P&L: growth that costs more in support than it returns in gross profit is not growth. It is purchasing volume.

The arithmetic survived. The calendar and the channel assumptions did not.

Built on the ADMADS Engine: sourced, dated evidence, with its limits stated. How these pieces are evidenced.

Nobody Decided to Make the Food Worse

A savings target is committed before anyone has named where it will come from. Every taste test that follows is run against last year's recipe, never the original.

You know the meeting. Next year's plan comes down, and inside it is a cost-savings number the business has already committed to. Nobody in the room can yet tell you which costs it will come out of. That is the next six months of work, handed to people who will be measured on finding it.

The person in the chair is a brand general manager, or the R&D lead sitting beside them. They have a contribution line to hit and four levers to hit it with, and three of those levers are defended by someone. This piece is about the one that is not.

Three levers have defenders. One doesn't.

Price is the most visible lever, so it is the most constrained. Key price points are thresholds retailers and shoppers both defend, and a private label sits one shelf over, waiting. Trade spend — the money paid to retailers for shelf space, displays and deals — cannot be cut without risking the distribution the whole business stands on. A&P, the marketing budget, can be cut and regularly is; it is the standard in-year rescue for a contribution target.

And COGS can be worked quietly, inside the product, where no line on the package announces that anything changed.

Pressure in a system flows to the levers with no defender. Marketing is one. The recipe is the other.

The savings number comes before the savings

Large food and household companies build a productivity target into the plan before anyone has named where it will come from. The number is committed first. Teams go and find it.

This is not an inference about how companies behave in private. It is disclosed to shareholders.

Some of what gets found is genuinely free. Automation, less waste, smarter logistics, procurement scale — nobody's dinner gets worse because a factory got more efficient.

The rest is found in the product. The industry's term is value engineering: re-engineering the product to deliver the same apparent value at lower cost. A cheaper oil in place of the premium one. Less of the characterizing ingredient — the cheese, the cocoa, the fruit, the meat — and more of the inexpensive extenders: water, starch, air. A thinner coating. A smaller piece. Or the close cousin shoppers have learned to name: a slightly smaller pack at the same price, shrinkflation, which is a price increase wearing a COGS costume.

Each change, on its own, is small. That is not a flaw in the process. It is the process.

Every test is run against last year's recipe

Value-engineered changes are validated by consumer testing, and the standard is a parity test: the new version must be not significantly worse than the current product.

Read that benchmark again. The current product. Not the original. Not the recipe that built the brand.

No individual step fails a test, no individual manager approves "worse," and every decision is locally rational: hold the price point, hit the savings target, pass the test.

Two things make the ratchet turn further than instinct suggests.

The step is large. A fifth of the salt, gone, undetected, in one move. And the method is not a trade secret — the paper's own introduction states the general case plainly: threshold methods are "typically used in ingredient replacement for cost savings, regulatory requirements, or product modification, where the change must not be evident to the consumer." A 2012 paper in the Journal of Sensory Studies is titled "Determining Sequential Difference Thresholds for Sodium Reduction." Sequential. The procedure for repeating it has a name and a citation trail.

The test is built to pass. A parity test returning "no significant difference" has not shown the two products are the same. It failed to show they are different, which is a different statement. A small panel, a noisy attribute or an underpowered design will return that result routinely when a real difference is sitting there. The weaker the test, the easier it clears — and nobody in the chain is rewarded for running one strong enough to fail.

The saving lands this year. The erosion lands on someone else's watch.

A COGS saving lands immediately, in full, in this year's contribution, credited precisely to the team that delivered it. The cost of a slightly worse product lands slowly, over years, in the base business — a household here and there buying less often, then not at all, attributed to the category, to competition, to anything diffuse. Brand managers rotate every couple of years. The saving is booked on one watch; the erosion is discovered on another.

Any system produces more of what it credits promptly and less of what it debits vaguely. This one is no different. And when the base softens, the standard responses make it worse: more trade promotion to hold volume, which squeezes margin further; a marketing cut to rescue contribution; and next year's productivity target, still in the plan, waiting to be found somewhere.

The product ends up funding its own decline.

The Number We Got Wrong, and How Long It Took Us to Notice

We published a perfect forecasting record for eight days. An audit of our own files found a prediction marked unresolved and counted as a win — and a much larger problem underneath it.

Every claim a research firm makes about itself has the same weakness: the firm keeps the scoreboard. We have been publishing ours for weeks — a record of predictions registered before the data was pulled, scored either way, misses included. It was the strongest line we had, and it was on a live page on this site.

This week we rebuilt it from source. It was wrong twice, and the second way was worse than the first.

This piece is for anyone weighing a vendor’s track record — ours or anyone else’s. The question is not whether the number is impressive. It is whether the number was derived from something you can check, by someone who would have published it either way.

The small half first. In October we ran a demand study on a recording artist and registered four predictions before pulling anything. One of them — whether a song or lyric term would outscore her name — could not be measured, because the instrument had been specified in a way that removed the term from its own count. The calibration record says so plainly: not measurable, unresolved.

A summary written the same day, in a different file, rolled the running record forward and counted that prediction as a win. Fifteen for fifteen became seventeen for seventeen. The correct figure was sixteen for sixteen.

Nobody recomputed it, because it arrived already computed. Over the following eight days that extra win traveled into our research design standard, into our publishing standard, onto a live study page, and into two articles drafted this week. Every document that repeated it cited a real source. The source had itself copied the figure.

Correcting one win would have been easy. The reason it took a rebuild is what the audit found next.

Our method separates predictions by where they came from: a prediction grounded in an earlier measurement with the same instrument is one thing, and a prediction grounded in something we merely believed is another. That split is the entire claim. Without it, a track record is just a batting average.

When we swept every project from the window, we found that 26 of roughly 33 had registered their predictions with a probability band, a falsifier and a result — and no basis label at all. Those predictions could never have entered the record, because the record is organized by a field they do not have.

So the published figure was not only one too high. It described about a fifth of the work and read as though it described all of it.

Rebuilt from the 48 predictions that carry an explicit basis: predictions grounded in a measurement held 14 of 19 times, or 72%. Predictions grounded in belief held 18 of 33, or 52%. The split survives. It is about twenty points. It is an edge, not a guarantee, and it is a great deal less impressive than the thing it replaces. Measured predictions do miss. Beliefs land about half the time, which is roughly what you would expect from beliefs.

Two things about the corrected figure are worse than they look, and both are on the ledger page rather than buried here.

One project supplies more than half the sample. A single brief registered 26 of the 48 predictions. Remove it and the gap widens to 86% against 42%. Read it alone and the gap is 64% against 62% — no measurable edge at all. That project registered 22 predictions in 73 minutes. Whether registering predictions at that speed degrades the care taken over labeling them is an open question on the page, not an explanation we have earned.

And one prediction held on a band it could not have failed. We predicted a search figure would land within 10% of the prior year, on an instrument whose own resolution at that value is about 10%. It held. It could not have done anything else informative. That is recorded as a hit with the defect stated beside it, because quietly dropping an inconvenient row is the thing this whole exercise exists to prevent.

We are publishing the weaker number for the same reason we published the stronger one. A perfect record is the least believable kind — it invites the suspicion that the scoring did the work. A record with its own audit attached is harder to dismiss, because the embarrassing parts are in it.

The corrected ledger is live, with all 48 predictions, their bands, their bases and their outcomes: every call we registered, including the ones we got wrong.

Twelve Days Nobody Wanted. An entire industry vacated the same window.

Grand Theft Auto VI is dated November 19, 2026. Two published calendars contain exactly one title in the twelve days around it — and the twelve days immediately before are the busiest stretch of the autumn.

You know the date meeting. The calendar is up on the screen, the build is three weeks short of where it needs to be, and the question is whether to move. The producer wants the time. Marketing wants a week where nothing else is shouting. And there, in the middle of the quarter, is a stretch of days with nothing on it at all.

An empty window looks like a gift. It costs nothing to take and appears to hand you a whole audience’s attention for a week.

What nobody in the room asks is why it is empty.

Twelve days, one title

Grand Theft Auto VI is dated November 19, 2026, on Rockstar Games’ own newswire. That source is primary and the date has moved twice before, so every figure here is anchored to it rather than to anyone’s memory of it.

Two published release calendars carry enough November entries to support a density comparison. Both contain exactly one dated title in the twelve days from November 13 to November 24. In both, that title is Grand Theft Auto VI. Neither contains a single title sharing its release date.

The reading that suggests itself is that late November is a quiet time of year. It is not.

Video Games Chronicle’s list is far sparser and shows the same shape at a different scale: 0.58 dated titles a day in the first half of November against 0.08 in the second, seven to one. Driffle, sparser still at four titles for the month, cannot support a rate and is reported as unable rather than folded in. It agrees on the thing all three agree on, which is that nothing else is dated to November 19. Six clear days sit before the launch in both calendars; eleven follow it in Video Games Chronicle, five in GameSpot.

Three calendars, one month, three answers

Ask three published calendars how many games come out in November 2026 and you get four, eight, and thirty-one.

None of them is wrong. Each admits a different set of indie releases, platform ports and remasters, and the inclusion rule is undisclosed in all three cases. “Number of games released in November” turns out to be an editorial filter wearing the clothes of a measurement — which is why nothing here compares one calendar’s count to another’s. Each is measured against itself, and the finding is only what survives in all of them.

The panel had one title in it

What this was meant to be was a displacement measurement: a panel of same-genre titles releasing into the window, an off-genre panel beside it, weekly demand for each against its own eight-week baseline. The brief registered a stopping rule before any retrieval — fewer than four titles in either panel and the pass stops and reports that the panel could not be built.

The same-genre panel has one title in it, and that title is Grand Theft Auto VI. There is no displacement to measure because there is nothing left in the window to displace. All four registered predictions are recorded as unscorable rather than quietly replaced with something the data would support.

The registered budget was fifty cents and none of it was spent. What made the study impossible was legible in a free calendar before a single paid row was pulled, and the lesson went into the register in those terms: a brief that registers a panel should register the panel’s existence as a precondition.

An empty week is a decision somebody else already made

The seat this was written for is a publisher’s release-date owner with two moves available: hold to the next quarter and absorb the delay, or ship into the window and absorb whatever displacement is real.

That question is not open. Everyone else holding a comparable title has already answered it unanimously, and the answer is visible without a budget or a data vendor. A publisher that believes it has found an uncontested week has found a week an entire industry independently vacated, and those are different objects.

The move that follows is to work out what the people who left knew — and sometimes it is nothing, and they all flinched at the same shadow. This window resolves on November 19, 2026.

The honest limits

This measures a published schedule, not shipped products. Dates move, and some undated titles will land inside the corridor. Undated titles are excluded entirely by all three sources, which give no way to know how many exist. Three calendars are not the universe of calendars, and all three are commercial publications with inclusion incentives they do not publish.

No causal claim is made anywhere here. The corridor is consistent with publishers avoiding Grand Theft Auto VI, and it is also consistent with a Thanksgiving-week effect: US Thanksgiving falls on November 26 in 2026, inside the thin stretch. The first half of November controls for that partly and not wholly, so it stands as an uncontrolled confound. The test that would settle it is the same three calendars read for a year in which the holiday and a generational launch do not coincide, and it should be run before this finding is leaned on hard.

There is no historical control, because comparing November 2026 to November 2025 would compare two editorial filters and reintroduce the error the within-month comparison exists to avoid. And no publisher was asked anything, so the qualitative half of this read is missing and the read is not finished.

The Screens Were Spoken For. Disney built its own format rather than go without them.

Two wide releases open on December 18, 2026, and one of them holds the IMAX screens. That was settled ten months before the date, by which point the only decision left to the other studio was what to build instead.

You know the date meeting. A release date goes up on the board eighteen months out, somebody says it is a strong date, and the room agrees. Then the conversation moves to cast availability and a marketing beat in the spring, and the premium-format question gets left where it usually gets left — for later, when the picture is further along and there is something to show.

The trouble with later is that the screens are not allocated later. They are allocated by someone else, on their own calendar, and that calendar runs ahead of yours.

Two films, one December Friday, one chain of screens

Avengers: Doomsday and Dune: Part Three both open in the United States on Friday, December 18, 2026. IMAX is a single chain of screens and it can only run one of them on an opening weekend.

It is running Dune. IMAX said so itself in an investor presentation that two outlets read and reported on February 26, 2026 — GamesRadar, citing slide 13 of the deck, and Gizmodo the same day. Ten months before the release.

On every measure of demand we took, Avengers is the larger film and it is not close: 95.3% of head-term search, 96.8% of trailer searches, 86.1% of official-trailer comments, 64.9% of trailer views. The narrowest gap is ticket searches at 73.2%, 90,500 to 33,100, and that is the measure sitting closest to someone actually buying.

That is a measurement of what the screens were worth, and it is made of behavior rather than projection. Losing one chain for one opening weekend was expensive enough to justify standing up a certification program and recruiting thousands of exhibitor screens into it inside a quarter.

The window closed in the winter

The answer engines recommend the smaller film

We put two questions to ChatGPT and Perplexity, three identical calls each. Asked which film is bigger, every instrument says Avengers. Asked which one to go and see, Perplexity named Dune: Part Three all three times and gave the premium-format reason every time. ChatGPT declined to choose in two of three, and in the third flagged the same constraint before calling Avengers the more accessible of the two.

The demand data and the recommendation layer disagree because they answer different questions. One is about how many people want a thing. The other is about which experience is available on the night.

There is a crack in that recommendation. Perplexity stated the exclusivity ran “at least its first three weeks” all three times, and the article it cited contains no duration at all. Three weeks traces back to the January post on X; the claim that traces to IMAX’s own presentation describes the opening weekend. The engines took the weaker number and stated it with confidence.

The honest limits, including one of ours

This study’s own pre-registered brief had the release date wrong. It said December 19; all six model answers said December 18, and the models were right, because December 18 is a Friday and US wide releases open on Fridays. The error came from a trade article read once and not checked against a second source — the rule this practice applies to everything except, that day, its own premise.

The second miss is larger. The brief named the right seat, whoever allocates premium-format screens, and then registered that decision as live. It was not. The allocation had resolved before we measured anything, and the engines were reporting an outcome rather than forecasting one. Picking the right seat and assuming an open decision is half a hit, and the half that missed is the half that decides whether a study is useful.

Nothing here observes a ticket sold, and the allocation itself is reported by outlets reading a deck we did not retrieve. Keyword volumes are Google’s advertising bands rather than counts, so 95.3% and 73.2% are band ratios and should be read as ordering. The trailer comparison is one official trailer per title published four months apart, which cuts toward Dune and makes 86.1% a floor. Both engines were read at the platform-default tier, on one day, from one location, and the brief registered a negative control that was never appointed because no true negative was available.

The prediction on the research page is registered and dated: Avengers: Doomsday takes the larger three-day domestic gross, at 65%, against 30% for Dune: Part Three and 5% for the two landing within 5% of each other. It resolves on Monday, December 21, 2026 from the reported three-day domestic figures, and the result is published whichever way it goes.

Built on the ADMADS Engine: sourced, dated evidence, with its limits stated. How these pieces are evidenced.

Hershey and Mondelez Each Lost Eleven Points. Costs did not fall after 2022. They split.

Commodity prices came off their peak, and every pricing conversation since has assumed the savings are sitting somewhere. Across 42 food and household filers the savings are not there — and the commodity data says why.

You know the meeting. Somebody from finance has a chart of input costs on the screen, and the line climbs through 2021, peaks in 2022 and then comes down. It is a real chart and the numbers in it are right. The question that follows is the one you have been dreading since the agenda went out: why the margin has not come back with it.

You took price twice. You held it. The relief everybody read about arrived for somebody. The honest answer is that you do not know which of your inputs moved, or by how much, because the number you have been handed is a category average and the decision in front of you is about a specific jar of a specific thing.

The savings are not in the filings

Forty-two food and household companies with a billion dollars or more of 2024 revenue file their numbers with the SEC, and those filings are free to read. Computed the same way for all of them — revenue minus cost of goods, over revenue — the median gross margin across the group has gone down since 2019, not up.

The obvious objection is that this is the economy rather than the category. So the same calculation ran against fifteen filers from outside food and household, drawn as a control before the category numbers were looked at. They went the other way.

That opposition is why this is a finding about food and household rather than about 2026. Two groups, one instrument, one day, opposite directions.

Costs did not fall. They split.

Every kilogram of everything imported into the United States is recorded with a value attached, which makes it possible to ask what happened to an input between the 2022 peak and 2024 without asking anybody’s opinion about it.

The oils came down hard — soybean oil 27 percent, palm oil 25 — and aluminum plate, a packaging line rather than a food one, fell 16. If those are your inputs, the relief was real and it arrived on schedule.

The tree crops went the other way in the same two years. Cocoa beans doubled. Chocolate and cocoa preparations rose 26 percent on top of the 20 percent they had added between 2019 and 2022.

A median taken across companies buying from both sides of that split lands near zero and describes nobody. It averages a company whose costs collapsed against one whose costs doubled, and the number it produces is true of neither.

The spread between companies is fifty times the movement over time

Across the same forty-two filers, the tenth percentile of gross margin sits at 13.8 percent and the ninetieth at 61.2 — a spread of 47.3 points between companies in one year, against a median movement of 0.92 points over five years.

A brand lead benchmarked against a category margin is being measured against a number fifty times smaller than the distance between the companies it was built from. Where you sit is a fact about your company. Which way you moved is a fact about your category.

For the price decision that resolves into something usable. Weighted toward the inputs that came down, holding price is a recovery you have already earned and the burden sits on anyone arguing for an increase. Weighted toward the ones that went up, absorbing another year funds somebody else’s harvest, and the increase has to be sized against the input rather than against last year’s increase.

Two of four predictions were wrong, and the leading one was mine

Four explanations were written down and scored before any data was pulled. The leading one, at 55 to 80 percent, was that these companies kept most of the input-cost decline and are sitting on widened margins. It is dead: margin compressed, and cost of goods rose as a share of revenue by the same amount it fell. The second, that they passed the savings through to shelf, is dead for the same reason.

What survived was the candidate written as the dull alternative — that input costs never fell for these filers in the first place — and the commodity split makes it sharper than it was written.

The falsifier attached to the leading prediction never fired, which is a defect in how it was written rather than a pass. It joined two movements together and tested only the combination, so the prediction failed in a way its own test could not see.

The limits on the rest of it are sharper than the headline. An import unit value is a national average across every importer, before contracts, hedging, grade and freight — it is not any company’s cost line, and no cost line was decomposed by input here. The commodity split and the filer split are consistent with each other, and that is all this can say.

No filing prose was read, so there is no management explanation and no split between price and volume. The control is fifteen filers sampled rather than a census, and twelve filers of sixty-nine produced no usable comparison and are published by name. The consumer half — what any of this did to shelf prices — is missing, which means the read is not finished.

The Chips Changed Address. They did not change how concentrated they are.

A de-risking program moved US integrated-circuit sourcing off its largest supplier and onto a different one. Across three semiconductor trade lines the same five years give three different answers, and only one of them looks like diversification.

You know the slide. There is a map on it, and the map has fewer pins in one country than it had three years ago, and the heading says something about resilience. Somebody earned that slide. It took two years, a qualification program, a second source for parts that had one, and a travel budget nobody enjoyed spending.

The question underneath it is simpler than the map, and the map is not built to answer it. Did the exposure go away, or did it move?

The top three held almost exactly what they held before

US imports of integrated circuits ran $33.09bn in 2019 and $40.39bn in 2024, across 114 trading partners in both years. That is the line most people mean when they say chips, and it is the line most de-risking programs were pointed at.

Over those five years the share held by the three largest partners went from 64.8% to 62.3%.

Read that number alone and the conclusion is that nothing happened. That conclusion is wrong too, and the way it is wrong is the whole point of the exercise.

Malaysia left and Taiwan arrived

The same question returns a different verdict per trade line

Integrated circuits relabeled. Discrete devices did something else entirely: Vietnam gained 9.6 points, Thailand 9.3, India 6.0 and Cambodia 5.9, while Malaysia lost 13.3 and China lost 5.8. More countries participating, by a wide margin — and the top three finished holding more than they started with, 55.1% against 52.3%. Wider participation, same concentration, and those two facts are not in conflict once you look at who the three are.

Manufacturing machines are the only line that behaved the way the diversification story describes. Japan and the Netherlands each gave up 14.7 points, the gains spread across Malaysia, Singapore, Korea and Germany with no single winner above 6.3, and the top three fell from 77.9% to 54.2%.

So the answer to “did the semiconductor supply chain diversify” is yes for the machines that make chips, no for the chips themselves, and the opposite of yes for discrete devices. Four explanations were written down before the data was pulled, and each one turned out to be true of at least one line and none was true of all three.

Taiwan is filed under a name you probably excluded

UN Comtrade does not report Taiwan as Taiwan. It reports it as Other Asia, nes — not elsewhere specified.

An analyst who filters that dataset by country name, as anyone would, gets a clean-looking table in which Taiwan does not appear in US semiconductor imports at all. The correct reading is that Taiwan is the largest single source of US integrated circuits in 2024 and the biggest share mover of the five-year window. The row is sitting there. It is just not wearing the name anybody searches for.

This is the kind of defect that never announces itself. The table parses, the percentages sum, the chart renders, and the single most consequential line in the dataset is missing from it.

Supplier margins went up, and Intel's went down

Across 41 semiconductor filers, median gross margin rose 2.22 points between FY2019 and the latest filed year, with 28 of the 41 expanding. Taking the four largest AI-exposed names out of the panel moves that median to 1.83 points, so the rise is not one company carrying the field. The 23 hardware and systems filers moved 1.41 points on the same measure.

For a procurement seat that line matters more than it looks. A premium being charged for diversified sourcing is not funding a margin squeeze at the supplier, because over this window there was not one.

The honest limits

Trade data records a border crossing, not a fab. A chip packaged in one country from a wafer made in another is credited entirely to the packager, which cuts directly against reading any of these shares as manufacturing location. That is the largest limit on the whole study and it does not have a workaround in public data.

Share is not volume and not capacity — a partner’s share can move because it grew or because somebody else shrank. The endpoints are 2019 and 2024, two data points, and the path between them is not described. Nothing here reads a line of management commentary from any filing, because the document host is unreachable from the workspace that ran the pull.

One registered prediction died outright: that the period would show real diversification. It is dead for integrated circuits, where the falsifier fired, and it holds for manufacturing machines. Fourteen filer-years were rejected by a plausibility band tuned for goods producers, which misfires on IP licensors above its ceiling and on loss-making producers below its floor. All fourteen are published by name.

No procurement lead was interviewed for any of this. It is an outside read of public filings and public trade records, and it is missing the half that would come from asking the people who signed the contracts.

Euphoria Opened Twice as Big. Three months later it was nine times smaller.

A renewal gets decided on an opening. Across thirteen 2026 streaming titles the opening month predicts rank and not retention — and at the top of the slate, where the expensive decisions are, it inverts inside a quarter.

You know the slate review. Twelve titles, twelve slides, and the same chart on every slide. Opening week. Opening month. The spike, the annotation, the arrow. Somebody has built it in the same template twelve times because that is the number that arrives first and arrives clean, and the meeting has to happen on a Tuesday whether or not the audience has finished deciding.

Then the renewal conversation starts, and it is a different conversation than the chart is equipped for. The question is not how big the launch was. The question is whether anyone is still there.

The opening is a marketing readout

A show’s opening month is the month its trailer ran, its cast did press, its key art went up on the home screen and its title sat at the top of the row. Those are things the service did. They are expensive, they are deliberate, and they work — which is the problem, because a number produced largely by your own spend is a receipt for the spend rather than a reading of the audience.

The tail is the part the service did not buy. Three months after release the promotion is over, the home-screen row has moved on, and whatever demand is left is demand somebody kept on their own. That is the quantity a renewal is actually about, and almost nobody puts it on the slide, because it takes three months to exist and the meeting is on Tuesday.

Twenty 2026 titles went into the measurement and thirteen had a full three-month window by the time it was taken. Four were thrown out because their names belonged to something older — a term that was already carrying 16,792 searches a month in 2019, seven years before the series existed, is measuring a restaurant group and a novel series rather than a show.

The thirteen that survived do not decay at a common rate. They do not decay at rates near each other. The best held 44.9% of its opening at three months and the worst held 2.2%, and both of those are shows a service was proud of.

Euphoria and The Pitt changed places

The tempting read is that the correlation is just weak. It is not. Across all thirteen titles the rank at three months tracks the rank at release at 0.841 — high enough that if you saw it on a slide you would stop worrying.

That number is carried by the bottom of the panel, where small titles stay small and the order never moves. At the top it reshuffles completely: Euphoria goes from first to fourth, The Pitt from third to first, and a title nobody was arguing about climbs past both of them. The correlation is strongest exactly where the stakes are lowest.

Two different mistakes, and they are not symmetrical

Renewing on a loud opening that did not hold means committing a season of budget to an audience that has already left. The money is spent before the evidence that it was the wrong call becomes available, which is the expensive direction.

Canceling a quiet opening that would have held means handing a durable title to a competitor’s licensing desk, and the cost of that one does not appear on any report you will read. It shows up as a show that did well somewhere else.

Both mistakes come out of the same habit — reading a level as if it were a slope. The opening is a level. Retention is a slope. They are different quantities, they are both knowable, and only one of them is in the room by default.

The honest limits

Search volume is interest, not viewing. Nothing in this measurement observes a minute watched, and a service’s own completion data would answer the question better than anything here — it is just not public, and this is. A renewal is decided on viewing, completion and cost, and this measures one input to it and does not pretend otherwise.

Google bands and rounds the figures it reports, which is why The Pitt reads 44.9% at both two and three months: the underlying value is the same band twice, not a plateau anyone can vouch for. The panel is declared rather than drawn from a register, eleven of the thirteen titles are returning seasons with audiences they already had, and one title of the twenty returned no series at all and is recorded as absent rather than as zero.

Two of the four predictions registered before the data was pulled came out wrong, and a defect in how two others were written — they were registered as alternatives and both turned out to hold — is published with the rest of it.

Built on the ADMADS Engine: sourced, dated evidence, with its limits stated. How these pieces are evidenced.

Rockstar Is Not the Source for Rockstar’s Release Date

One question, three machines, one minute: twenty-seven cited domains and not one shared by any two of them. If your AI visibility number came from a single call, it is a reading, not a measurement.

Somewhere in the next six weeks, in a quarterly review, a slide will go up with a single number on it describing how often a brand appears in AI answers. Someone will have paid for that number. Nobody in the room will ask how many times the question was asked.

One budget line, or two

The person this matters to is whoever has to defend next year's marketing budget, which for most calendar-year companies is being set right now. The ask is a separate line for AI answer visibility. The refusal is almost always the same sentence: answer engines read the web, Google ranks the web, so winning search wins both. One line, one owner, done.

It is a reasonable argument. It is also, as far as we can tell, untested by almost everyone making it. So on October 24 we tested it on one question, in one city, inside a single minute, with every row published.

We picked a question with no ambiguity in the answer. Grand Theft Auto VI ships on November 19, 2026. Rockstar has said so. It is on the store pages. Some phrasing of when does it come out is searched well over a million times a month, which makes it one of the least obscure facts on the internet.

Ask the same machine the same thing three times

ChatGPT got the date right all three times. That part is unremarkable and worth saying plainly, because the first version of this piece claimed otherwise and was wrong.

What moved was where it said the date came from. Across three identical calls it cited seven different domains, and exactly one of them appeared in all three answers. Six of seven were present in some calls and absent from others. A brand measuring its citation share from a single call would have been reading one draw from a shuffled deck.

Perplexity, asked the same string three times, returned the same ten domains in the same order, three times out of three. That is the fourth separate occasion we have measured that split. It is a measurement of these questions on these days, not a league table, and we would not publish it as one.

Then ask three different machines the same thing

This is the part that bears on the budget argument, and it is blunter than we expected.

Read the last row twice. Google was less consistent with itself across two phrasings of one question than Perplexity was across three repeats of the same string. Same engine, same city, same minute, and rewording the question changed five of six results.

Who the machines actually asked

We registered, before the pull, which domains we expected to see: Rockstar's own site, Take-Two's, IGN, GameSpot, Eurogamer. The obvious authorities on a games release date.

Across three ChatGPT calls, that list returned nothing. Of its seven cited domains, exactly one — PC Gamer — was mainstream press. The rest were fan-built wikis and single-franchise sites. On the third call, all four citations were sites that exist to answer questions about this one unreleased game, and nothing else.

Rockstar Games was cited zero times out of three on a question about Rockstar's own release date.

CULT+MATH, run 3063e451, October 24, 2026

Perplexity's top two positions went to the same kind of site, ahead of the BBC, IGN, CNET and USA Today, all of which it also cited further down. Six of the seventeen domains the two engines cited between them exist to answer about this one product.

We are going to resist the obvious conclusion. These sites were cited; nothing in the run measures why, and a correlation between being single-subject and being cited is not a mechanism. One query, one day. It is enough to change what we test next. It is not enough to tell anyone to restructure a website.

What this evidence cannot say

Organic results are a minority of the Google page. We requested depth ten and got six and seven organic positions; the rest of each page was an AI overview, a video carousel and paid units. A statement about organic overlap is not a statement about what a person sees.

Both engines were read at the platform-default budget tier the API serves, which is not the model a consumer is given. Every AI visibility measurement we have found, including our own, has this limit, and most do not mention it.

One query, one city, one day, one product. Nothing here observes a visit, a purchase, or a person. And the strongest result in the run — the zero overlap — came from a comparison we had already run badly earlier the same day, which means we had seen a number before we wrote down what we expected. The study page says so, in the section where the prediction is recorded.

The question underneath all of this is not about AI. It is the oldest one in measurement: how many times did you look before you wrote the number down? Every finding in this piece exists because the answer was three instead of one. That is not a sophisticated technique. It is the cheapest possible rigor, and almost nothing published in this category currently clears it.

Every row behind this piece — all twenty-seven domains, the per-call presence, the advertising series, the search demand and the comment vocabulary — is published and filterable in the standing study at cultmath.com/research/games/gta, which is re-run weekly and corrected in place, dated. It currently carries one retraction, made in the same pass that produced the measurement that overturned it.

Built on the ADMADS Engine: sourced, dated evidence, with its limits stated. How these pieces are evidenced.

Eighteen Ways to Ask One Question

In September I argued that an AI visibility score is mostly a property of the tracker. The load-bearing evidence was somebody else’s preprint. This is my own, and it agrees on the direction, disagrees on the shape, and turned up a failure nobody has named.

Six weeks ago I published a piece arguing that the number on an AI visibility dashboard is mostly a property of the tracker rather than of your business. The argument rested on a study by Will Jack and colleagues — roughly 6,000 paraphrase trials against 6,000 same-prompt controls, showing that rephrasing a buying question in a second natural way produced a largely different set of recommended brands.

At the bottom of that piece, under the honest-limits heading, I wrote that the load-bearing source was a preprint. That was true and it was uncomfortable. A measurement argument that depends entirely on somebody else’s measurement is a book report.

So this week I built the smaller version of it on my own instrument and ran it. Three questions, six phrasings each, eighteen measurements, about eleven cents. What follows agrees with the preprint on the direction, disagrees with it on the shape, and found one thing that is not in it at all.

First, the part that looked stable

The seeds were not invented. On October 21 I ran fifteen questions a grocery buyer or distributor actually asks, sourced from retailer supplier pages, broker line-review guides and published buyer interviews. Twelve of the fifteen were led by a site almost nobody has heard of: a marketing agency, two software tools built for packaged-goods brands, a broker, a founder community.

Three of those incumbents became the seeds here, chosen because each already held first position on a real question. There was something to displace.

Asked again, in the same words, each of the three held. That is what stability looks like before you test it.

Then I changed the words

Six variations per seed, with the kinds of variation fixed before the test ran so that a failure could be traced to a type of rewording rather than to one awkward sentence: the exact original, a synonym swap, a grammatical reframe, added first-person context, a stripped-down short form, and the colloquial version a person would actually type.

This is where my result parts company with the preprint. Jack et al. found a gradient: a cosmetic rewording cost less than adding a buyer’s constraint. I found no gradient at all. Swapping one content word for a synonym did no better than writing the question the way a tired founder types it. The exact question behaves like a cliff edge, not a slope.

There is a second reading in that table and it matters more than the headline. Count presence rather than position and the picture improves sharply: the incumbent was still cited somewhere in the answer on an average of 4.67 of 6 rewordings, against 2.67 of 6 for first place.

Being in the answer survives rewording. Being the answer does not. The preprint measures overlap between recommended sets, which collapses those two into one number. Separating them changes what a content program is for.

And then one question left the category

One of the three seeds asked how grocery buyers use total distribution points — the measure of how widely a brand is already sold, weighted by the size of the stores selling it. Five phrasings returned grocery sources.

The sixth returned hp.com in first position.

"What is TDP and why does my buyer keep bringing it up?"

The colloquial variant. TDP in computing is Thermal Design Power.

Three letters mean two things. The assistant chose the meaning with more written behind it, answered confidently about processor heat, and never came back to groceries. Nothing errored. The row looks exactly like the other seventeen.

That collision was written into the test before it ran — the row carries a note, recorded at build time, saying the acronym means something else in computing and that a computing result there should be read as a collision rather than as a finding about grocery. It collided. Which is why it is a result rather than an embarrassment.

Here is the part that should bother anyone buying one of these reports. The phrasing that failed was the one a real person would use. The five that held were mine: careful, complete, industry named, written by somebody who already knew the answer he wanted. The sixth is how a founder types after a meeting where a term came up three times and they did not want to admit they did not know it.

A tidy question list is a biased question list. It is weighted toward the phrasings that behave. The messy ones are the likeliest to leave your category, and the least likely to be on anyone’s list.

Where this agrees with September, and where it does not

It agrees that wording moves the score more than the business does. That was the September argument and it survives first-party contact.

It disagrees on the shape. The preprint found decay by semantic distance. I found a cliff. If that holds up, “how far did you rephrase it” is the wrong question and “did you use the exact words” is the right one.

It disagrees on stability under repetition. Jack et al. report same-prompt rerun overlap of roughly 0.50 to 0.61 on OpenAI and Anthropic models. Asking an identical question three times on Perplexity, I get the top seven positions identical three times out of three, with movement only at position eight. That is almost certainly a platform difference rather than a contradiction, and it means a single reading on one platform is more trustworthy than the preprint’s baseline would suggest, while a single reading on another may not be.

And it adds a failure mode the preprint does not contain. Paraphrase brittleness is a question of degree. A domain collision is categorical: the answer is not worse, it is about a different industry.

Where this argument is weak

Three seeds. The variation table rests on three observations per row, which is enough to see a cliff and not enough to measure its height. One platform, one day, one generation per row, on the documented budget-tier model rather than the one a paying buyer is served.

The incumbents are other people’s pages. Our own answer pages went live the day before and have no post-publication reading, so whether a new page behaves like an established one under rewording is untested here.

And the seed whose colloquial variant collided is the weakest of the three on every other measure. Excluding that row it still scores worst, so the finding survives the exclusion — but the headline average would be wrong to quote without saying so.

Finally: none of this says how many people ask any of these questions. No instrument I trust measures question volume inside an assistant, and the one that claims to is derived from Google’s related-questions box, which inherits the exact blind spot it would be used to test.

Rolex Has a Page for Jon Rahm. Ask the Wrong Way and Bob’s Watches Answers.

Rolex publishes a page about Jon Rahm on its own domain. Asked one way, a language model cited that page first, three times out of three — asked another way eleven minutes later, it cited a dealer in pre-owned watches and Rolex appeared nowhere.

The ambassador page goes live on a Tuesday. The photography has been signed off twice, the copy once more than that, the right watch is on the right wrist, and it sits on the house’s own domain where it belongs. Somebody posts the link in a channel. Three people react. Everyone moves to the next launch.

That page is now doing one of two things, and nobody in the building knows which. It is either the answer your buyer gets when they go looking — or it is a beautifully produced asset that the thing doing the answering has never opened.

On October 18, 2026 I asked a language model, with live web search switched on, a question a buyer might plausibly type: which watch brand is Jon Rahm an ambassador for? I ran it three times, because these models are not deterministic and one answer is an observation rather than a finding.

All three times, the first source it cited was bobswatches.com — a dealer in pre-owned Rolex watches. The second source varied: a golf title, a trade paper, once Rahm’s own site. Rolex was not cited once.

Rolex has a dedicated page for Jon Rahm. It is on rolex.com, in the section the house keeps for its Testimonees. It was live the whole time.

The winning phrasing did two things the losing one did not. It named the brand inside the question, and it used Rolex’s own word — the house calls its athletes Testimonees, not ambassadors. So I ran the other two combinations to find out which lever was doing the work.

Both of them were, independently. Naming the brand was enough on its own. Using the house’s own vocabulary was enough on its own, though it routed to Rolex’s newsroom rather than the main site. Only one of the four sentences lost, and it is the one a buyer is most likely to use.

Roger Federer made it stranger. Asked the open way, the model did not search at all — it answered from memory in thirteen words and cited nothing, three times out of three. Asked does Rolex sponsor Roger Federer?, it searched and put rolex.com first. For one of the most famous endorsements in sport, one phrasing produced no citation surface whatsoever.

If the story ended there it would be a trick about words. It does not.

That leaves two conditions, and both have to hold. The content has to exist on a domain you control, about that specific association. And the sentence has to reach it. Rolex had the first and was failing the second, which is why a rewrite fixed it in eleven minutes. TAG Heuer, for Elordi, appears not to have the first — so no sentence helps.

Across all 89 measurements that day, when the question named the brand, the brand’s own domain was cited in 82% of rows. When it did not, 42%. Questions about who founded a company returned zero owned citations across 17 rows, and in nine rows the model never searched at all.

You hold a roster of named associations — ambassadors, licensed products, signature lines — and every one of them is currently being answered by somebody. Here is how the six seats around the table read it.

• CEO. The association is an owned asset the house paid for, and if a reseller is the answer, the equity is being built in one place and collected in another. That pulls against the CFO, because the fix is unglamorous page work rather than a campaign with a launch date. Move: treat the roster of associations as inventory, and know which ones answer.
• CFO. The page is a sunk cost already on the books, and finding out whether it is reachable costs pennies per association against what the shoot cost. That pulls against the CRO, because it buys information rather than reach. Move: fund the audit, not the rebuild.
• CRO. A buyer asking the open question meets a dealer’s page instead of the house’s, at the exact moment they are deciding. That pulls against Legal, because a reseller ranking first is not a rights problem but it reads like one in a meeting. Move: map the phrasings buyers actually use before commissioning any new page.
• CPO. The fix is a page shaped like the question, not more pages — the asset exists, and its form is wrong for how it is being asked for. That pulls against the CEO’s instinct to commission something larger. Move: publish one page in the losing form and re-measure.
• Legal. A pre-owned dealer answering for Rolex is lawful and ordinary. The real exposure in this data was elsewhere: at another brand, a staging server was cited as the top source on a question about that brand’s own products. Move: find out which of your own hosts are publicly reachable before worrying about anyone else’s.
• Risk. The dependency is a retrieval behavior nobody controls, and it moved inside eleven minutes in this data. Move: measure on a cadence, never once.

The reconciliation. CFO and CPO want the cheap version, the CEO wants the version with a budget line, and Risk wants it repeated. The course that satisfies the most of them is to audit first and publish second — because until the phrasing map exists, nobody at that table can say which page is worth building, and the audit costs less than the meeting about it.

It cannot say anything about demand. Not one of the 89 measurements looked at willingness to pay, traffic, search volume or revenue. Being the answer and being chosen are different things, and this measures only the first. Anyone presenting a citation share as a demand figure is selling you a different variable.

The four-sentence grid is twelve rows at one brand. It is clean and it is small, and Rolex is unusual — it maintains a page per Testimonee and a separate newsroom host, which most houses do not.

Where a brand was not cited, I read that as the brand not having a page about that association. I did not verify it by hand. That inference is the weakest joint in the argument, and it is load-bearing.

And every question here is one I wrote. I do not know which phrasings real buyers use — and when the entire finding turns on phrasing, that is the gap that matters most.

Eleven predictions were registered with probability bands before any data was pulled. Four held. Six failed. One split.

The failures are the useful part. I predicted Rolex would be cited consistently because it has the page — it was not, until the sentence changed. I predicted creator-founded brands would score worst at owning their own story; they scored best of any class once I asked them a question they could win. I published five conclusions on October 18 and retracted all five the same day, each within hours, each because I read a mechanism into a result before testing the alternative. The timestamps are on the record.

Every figure, every cited domain and the full prediction ledger — including those five retractions — are published at the IP Association Map.

Built on the ADMADS Engine: sourced, dated evidence, with its limits stated. How these pieces are evidenced.

Related from Sound Decisions: How These Pieces Are Evidenced · Brand Strategy That Compiles · The IP Association Map

Brand Strategy That Compiles

The position document stops being prose and becomes a file — one that governs operations and can be read on the P&L.

Every brand strategy I have been handed in fourteen years arrived as prose. A paragraph of positioning. Three or four adjectives. A page on tone, a line about what we stand for. Then somebody reads it and works out what it means for the thing in front of them this afternoon, and two capable people reading the same page produce different work.

That was never carelessness. It was the medium. Words on a slide cannot be executed. They can only be interpreted, and interpretation drifts — which is a property of interpretation, not a failing of the people doing it.

That constraint has lifted, and the discipline has not noticed yet.

A model can read a sentence and act on it. Language that used to need a human in the loop to turn into action can now turn into action directly.

Everything that follows comes from that one fact. If a sentence can be acted on, a sentence can be a specification. And if positioning can be written as a specification, it stops living in a deck and starts living in a file.

Not automation. The automated-strategy version of this argument is wrong, and it is also the easiest version to sell, which is why it is the one in the market.

Software contributes two things. It deletes the performance — the adjectives that mean something different in every department, the criteria written after the results came in, the guideline nobody enforces. And it makes execution checkable, so a claim made in March can be tested a year later against the thing it actually rested on.

Three rows, and only one of them belongs to the machine.

The deck. Subjective brief, unenforced guideline, blind run.

The compiler. Sourced input, validated against rules you declared, refused or passed.

The judgment. What it means, what we stop doing, what we stand behind.

The compiler owns the middle row. Only the middle row.

The position is a markdown file. It sits in a repository beside the work it governs, with a version history, so a change to the strategy is a commit someone made on a date for a reason rather than a new deck that quietly supersedes the old one.

It does not describe the brand. It declares what the brand has to satisfy. The claims it is permitted to make and the ones it is not. The price floor, below which a promotion stops being a promotion. The audience parameters. The words that are load-bearing and the words that are forbidden. And precedence — which constraint wins when two of them collide, written before the collision rather than argued during it.

A pricing system cannot read "premium, approachable, innovative." It can read a floor.

That is the whole unlock. Once a constraint is declared rather than described, it can be handed to the systems that actually run the company. The price floor goes to the deal desk. The claims allowlist goes to the support macro library and the product page validator. The promotion rule sits at the approval gate. The service standard goes into the SLA.

Brand stops governing communications and starts governing any decision about what the company is willing to be known for — which is a much larger set of decisions, and most of them have never had a brand person anywhere near them.

The precedent is strong, and it is not from marketing. Security and compliance made this exact transition already: policy as a document became policy as code, and the annual audit became continuous controls. Nobody in those fields argues about whether the standard is being followed, because the standard is enforced at the point the work ships. Brand is close to the last strategic discipline still handing over prose and hoping.

That is the executable half. The verifiable half is the one almost nobody is building.

Once the position is declared precisely enough to be checked, it can be checked against the world rather than against the room. Not with a tracking study next year. With public data, continuously.

What people actually search for, and whether that is moving toward you or away. What claims a competitor's site carried last year and dropped this year. What a filing says about where the margin went. All of it published, free, and almost entirely unused by the people whose positions depend on it.

The question is our position still holding? stops being a research project with a budget and a six-week lead time. It becomes a query you can run on a Tuesday.

Put the two halves together and you get the thing brand strategy has never had: a positioning brief generated from evidence that knows when it has gone stale.

Every figure carries where it came from and when. Every gap is named in the section that needed it. Re-run the pipeline a quarter later and the brief updates itself — claims that still hold, hold; the ones that have decayed surface as changed, with the date and the source that says so.

This is what I have spent most of the last two years building rather than writing about. The ADMADS Engine is four layers. Sources — public filings, market demand data, claim archives, the category's own voice. Entities — a brand or a segment, written down once, so the thing being measured cannot quietly change. Recipes — a named read that expands an entity into evidence rows. Outputs — a brief carrying every figure, its provenance, and a block stating plainly what the evidence cannot support.

It produces briefs the way a build produces a binary: from declared inputs, reproducibly, with a record of what went in. A team does not discover eighteen months late that its position drifted — they watch it hold, or stop holding, while there is still time to act.

The reason the standards give is worth reading closely, because it is not dismissive. It is that the expenditure on building a brand cannot be reliably separated from the cost of developing the business as a whole. Not brand has no value. We cannot tell which spend built it.

The consequence inside a company is mechanical. Brand investment is expense. Much of what it competes against for the same capital can become an asset instead — plant and equipment as a rule, software and development spend where the tests are met. Treatment varies by expenditure and standard; brand is the case where the answer is always no. Expense hits this period's margin in full; a capitalized asset spreads across years. So in a soft quarter the fastest lever to protect margin is to cut expense, and because expense leaves no residual on the books, cutting it appears to destroy nothing.

Brand is structurally disadvantaged in the budget cycle. Not because finance does not believe in it — because the rules treat it as a cost while treating the alternatives as assets. Saying that out loud to a finance partner changes the conversation, because it concedes their frame is correct and moves the argument somewhere it can be won.

Inseparability is the objection. Lineage is the answer to it.

A declared constraint, the work validated against it, the spend that satisfied it, and the market signal that moved afterwards — each dated, each traceable backwards. A thread now runs from a brand decision to an operational change to something that shows up in the numbers.

This does not make an internally built brand capitalizable, and any piece claiming otherwise is selling something. The standard is the standard. What changes is the argument inside the building, and that was always the one that decided the budget. The CFO's real objection has never been disbelief. It has been I cannot tell what I am buying. That objection has an answer now.

Brand does appear on the P&L. It appears unlabeled, in four places.

Pricing power lands in gross margin — the price held while the category discounts. Unpaid demand lands in acquisition cost — demand arriving by name, the cheapest there is. Loyalty lands in lifetime value. Resilience lands in cashflow steadiness — dipping less in a bad quarter and recovering faster.

None has the word brand on it. All of them are the brand, converted into money, in the statement the CFO already reads.

The finance literature got there first, which is the useful part. Srivastava, Shervani and Fahey showed in 1998 that market-based assets raise shareholder value by accelerating cash flows, increasing them, reducing their volatility and vulnerability, and raising their residual value. Madden, Fehle and Fournier found that portfolios of strong brands outperformed the market on returns while carrying less risk. Strong brands make more money, more safely — and that comes from finance journals, not marketing ones.

What the constraint layer adds is the middle of the chain. Until now you could show the fingerprints and assert the brand caused them. Now you can show which constraint was declared, which work was held to it, and when the signal moved.

None of this decides anything.

The engine gathers and counts. It checks a claim against a constraint you declared, records where every figure came from, refuses to report a number it cannot support, and notices when a claim has changed. It cannot decide what the brand should stand for, which figure matters, what to do when the evidence is missing, or whose voice counts.

A tool that supplies judgment invites you to stop exercising it. The position is still a human argument. What changes is that the argument now has to survive contact with something that checks it.

That last one is the failure mode I would watch hardest, and it is subtler than Goodhart. Making brand executable means the executable parts get the attention. A price floor is checkable. Treat customers like adults is not. Encode the first and leave the second as prose, and in eighteen months the measurable constraints are immaculate while the ones that actually carried the meaning have quietly gone.

The fix is not to measure more. It is to keep the unmeasurable commitments in the same file, marked as unenforceable, so they are at least visible when somebody is deciding what to trade away.

Related from Sound Decisions: How These Pieces Are Evidenced · Nobody Pays You Extra for Being Honest · Build It Anyway

Liquid Death Moved the Money Down. The reason it gave is not the reason it works.

The brand is putting the majority of its 2027 media into retail networks and says awareness is no longer the problem. Its own tracking puts the widest gap at the awareness step, and what the money buys is a measurement.

The 2027 plan is on its third draft and somebody reads a line off their phone. A beverage brand everybody in the room admires has said out loud, at a conference, that retail media will be the majority of its spend next year. Not a slice. More than half. The room treats it as information, which it is, and then treats it as a template, which is where the trouble starts.

The brand is Liquid Death and the executive is Benoit Vatere, its chief media officer. He is among the more credible people in the category on this subject, and the move is defensible. The reason he gave is the part that will get repeated, and the part that does not survive contact with his own record.

That reasoning, as relayed: consideration happens as close to conversion as you can get, retail media is what in-store display used to be, and awareness is no longer the problem. On that account the money moves down because the top of the funnel is finished.

Six months earlier, the same man argued the other direction

In April 2026, talking to AdExchanger about how Liquid Death decides which tactics to kill, Vatere described the opposite direction of travel. He was explaining why he invests in bottom-of-funnel measurement, and his explanation was not that the bottom is where demand gets made.

"If I have a tight net at the bottom, I can feel a lot more confident pushing hard at the top."

Benoit Vatere, Chief Media Officer, Liquid Death — AdExchanger, April 8, 2026

Both statements can be true at once, and they are compatible in one way only — which changes what the decision is. If the net at the bottom licenses the spending at the top, money moving into retail media is money buying visibility into results. A brand that copies the share without building the net has bought the cost and none of the license.

Checking "awareness is no longer the problem"

Liquid Death's awareness in the United States moved from 26 percent to 34 percent across a year on Tracksuit's brand tracking, reaching 46 percent among 18-to-34s. Three percent of bottled-water drinkers name it as a preference.

Thirty-four percent is large for a brand founded in 2019. It is not a solved awareness problem: two thirds of the market has not heard of it.

The conversion between the steps says more. Liquid Death turns 41 percent of its awareness into consideration, against a competitor-set average of 50. Fiji converts 57. From consideration to preference it runs 21 percent against an average of 25.

So the step from awareness into consideration — the one the stated reason treats as finished — is where the brand trails its category by the widest margin. Nine points there. Four points at the step nearest conversion. The money is moving away from the weaker half of the funnel, not toward it.

The strongest counter-reading supports the move exactly as stated. A brand converting attention into purchase below its category average has a bottom-of-funnel problem, retail media sits on the bottom of the funnel, and spending there is spending on the leak. That argument holds. It is not the argument that was made. The argument that was made is that the top is done.

What he has actually been optimizing for

Read across his public statements for the twelve months before the Groceryshop remark and one problem keeps showing up. It is never funnel position.

Every row is about the ability to see and to move — an executive buying the capacity to run a controlled test on a short cycle, in a category where he has said there is no loyalty to coast on. That capacity currently exists in retail media networks with closed-loop sales data.

That makes the 2027 decision legible in a way the stated reason does not. He is concentrating budget where he can measure. Called a funnel reallocation, it is arguable. Called a measurement decision, it is close to obvious.

Deep, not wide — against a market going the other way

The second half of the Groceryshop remark got less attention. He is going deep with the big networks rather than wide, because splitting a budget across twenty of them dilutes the impact while the work per network stays the same.

The market is doing the reverse. In the February 2026 Skai and Stratably survey of 166 retail media advertisers, brands reported working with six networks on average and expected eleven by year end. Fifteen percent reported strong confidence in their measurement. Fifty-six percent named limited analytics capacity as the main obstacle to measuring incrementality. In NIQ's July 2026 analysis, 67 percent of chief marketing officers planned to increase retail media investment; 53 percent believed their networks measured adequately.

Those figures belong in the same sentence. Doubling the number of counterparties while five advertisers in six cannot confidently measure one of them is a decision to buy more of what you cannot read. Vatere's binding constraint is analytics capacity per network, and his answer is fewer networks. That reasoning does not depend on his awareness being solved, his brand being famous, or his category behaving like bottled water.

What this evidence cannot say

Four limits sit on everything above.

The 2027 figure is an intention, not a disclosure. It was said in a podcast conversation and relayed by one of that show's hosts. There is no published split and no trade-press account of the figure as of this writing. Read it as what he says he will do.

The brand-tracking figures come from a company that sells brand tracking, and the public write-up carries no sample size, field dates or methodology. The direction is probably right; the nine-point gap is not a precise quantity.

Nobody outside the company knows the current split, so there is no telling whether this is a large move or a small one described well.

The best evidence on brand-versus-activation weighting predates the channel. Binet and Field's analysis of 996 IPA case studies from 1980 to 2016 found 60 percent to brand building and 40 to activation as the profit-maximizing average, shifting toward 50/50 online, and in no sector analyzed did activation outperform brand building. Retail media did not exist in that databank — which makes the field's most-cited guidance a reason for care here rather than a rule.

Vatere is probably making the right call. He has said for a year that he will not spend where he cannot see, and retail media is where a beverage brand can currently see most clearly. Concentrating there, with a measurement platform attached and a monthly willingness to move the money again, is a coherent position reached from a stated constraint.

The reason he gave is a funnel argument, and funnel arguments travel better than measurement arguments. One sounds like strategy and the other sounds like plumbing. The brand that copies the share next year will take the sentence about consideration happening close to conversion, cut the upper-funnel work that produced the attention it has been converting, and keep buying across eleven networks because the eleven were already in the plan. The decision will have been copied. The thing that made it work will not.

Related from Sound Decisions: Most of Your Marketing Budget Is Rent · Chase Cut 395,000 Websites · Shelf Space Is Not Demand

Private Label Took the Record. It couldn't take the price.

Store brands reached the highest unit share they have ever held in the first half of 2026, and their dollar sales did not move. Nearly half of all private brand growth came from one channel. The threat is real, and it is not where the share deck says it is.

You know the slide. Store brand share, four years of it, climbing left to right. Somebody says the words "structural trade-down," and the room agrees, because the line on the slide agrees. By the end of the meeting there is a price investment on the table, or a value tier, or a promotional calendar heavier than last year's. Nobody has asked where on the shelf the share went, or what the shopper paid for it, because the slide does not carry either number.

It is the right chart and the wrong conclusion, and the gap between those two is expensive.

Here is what the same six months look like with the dollars added back. In the first half of 2026, store brands reached a record 23.8 percent of grocery unit share. Their unit sales rose 0.2 percent. Their dollar sales were flat. Their dollar share was 21.2 percent.

Read those four numbers as one sentence: private label moved more units than it ever has and did not take any more money for them.

A gap between unit share and dollar share is normal for private label and always has been. Store brands cost less per unit, so they index higher on units than on dollars by definition. That part is arithmetic, not news, and anyone selling you the gap as a discovery is selling you a price tag.

The news is the direction of travel inside it. A record volume position produced no revenue growth at all. Whatever private label gained in the cart over those six months, it gave back at the register — and it did that in a period when its competition was taking price rather than cutting it. Over the same six months national brands lost units, down 0.5 percent, and added 2.2 percent in dollars, a position with a shelf life of its own.

Circana, whose data the record is built on, describes the next phase in its own words rather than anyone else's.

The outlook for private label remains positive, though more balanced.

Sally Lyons Wyatt, Global EVP and Chief Advisor, Circana, March 31, 2026

The same research describes growth becoming more measured as the sector moves from acceleration into normalized cycles, and trade reporting on it notes intensifying competition from national brands as part of the reason. That is the firm that counts the units telling you the slope is flattening. It is not a reversal and nobody should sell it as one. It is a different thing, and a more actionable one: the trend you are being asked to fund a defense against is decelerating while the defense is being budgeted.

The second number does more damage to the slide than the first.

This is where the two sets of figures have to be handled carefully, because they are not the same universe. The 23.8 percent record and the flat dollars come from the PLMA midyear release on Circana grocery data for the six months ending June 14, 2026. The $330 billion and the club-channel finding come from Circana's own March 2026 research across a broader set of outlets, where private label reads 24 percent of units and 23 percent of dollars. Both are Circana. Neither is wrong. Quoting one against the other without naming which is how a confident wrong comparison gets made, and it is made often.

What survives in both universes is the shape: private label is winning units faster than dollars, and its growth is not evenly distributed across the store.

Trade-down read as a single structural force has one obvious answer, and the answer is price. Trade-down read as a volume win that cannot be priced, concentrated in particular channels, has a different answer, and price is not it.

The difference matters because the two cost different amounts and buy different things.

A price investment is permanent in practice. It resets the shopper's reference price, and reference prices do not go back up quietly. You can fund it for a quarter; you live with it for years.

A channel decision is reversible. Deciding to concede volume in the warehouse aisle and hold your price everywhere else is a decision you can unwind next year with nothing broken. It also has a name your finance partner already understands, which is mix.

And the two volumes are not worth the same money. This is the part most share decks are structurally unable to show. A unit lost to a club-pack store brand and a unit lost to the store brand beside you in the national grocery set are the same unit on the slide and different units in the P&L. One of them you may be delighted to lose.

Which raises the question the meeting never gets to. Not how do we close the gap — but which units are you actually willing to lose, and what do you have to be worth to the shoppers who stay?

Registered before writing, and the first two are live.

Flat dollars could be deflation rather than a price ceiling. If store brand unit growth continued while category prices fell, flat dollars is a pass-through, not evidence that private label cannot raise price. That is a real alternative explanation and the midyear release does not settle it. The check is private label price per unit against category price per unit over the same 26 weeks.

One half of 2026 is one half of 2026. A flat six months inside a multi-year climb is a data point, not a turn. The second-half figures and Circana's next annual read are the test, and they are not in yet.

Club concentration may not be your concentration. Nearly half of growth coming from club is a statement about the total market, not about your category. In a category where the store brand gains sit squarely in national grocery, the channel argument here does not apply to you and the price pressure is exactly as direct as it feels.

If premium private label is the real engine, this reads too favorably. Store brands moving into premium and functional tiers are not competing on price at all, and a dollar share that lags units tells you nothing about that fight. A premium store brand beside you is a positioning problem, and no price decision touches it.

Private label did not stop winning. It won the unit war in the first half of 2026 and took no more money for it, in one half of the store more than the others, while the firm counting the units called the next phase more measured.

None of which shows up on a share line climbing left to right. That is the problem with the chart, and it was never the chart's fault — it answers the question it was built to answer. Somebody in the room has to ask the other one.

Related from Sound Decisions: If You Take the Quality Out · A Third of the Shelf, Two Percent of the Growth · Distribution Doesn't Create Demand

Coca-Cola Ran It Again. The second time was the decision.

The first AI Christmas ad was called an eyesore and AI slop. A year later the company made another one, bigger, and said so on purpose. What it weighed is the same thing sitting in your meeting right now — and the evidence everyone reaches for says something neither side expects.

You know the meeting. Someone senior has seen a demo, and it was genuinely impressive, and the question arrives already answered: why are we not doing this. You have an objection and it is a real one. The humanity is the thing people are paying for. That is what separates us from the cheap version of us on the same shelf.

Then it is your turn to speak, and what you have is a conviction. What the other side has is a cost line. The meeting is going to be decided by whoever brings a number, and neither of you has one.

So the argument becomes an argument about taste, which is an argument you lose slowly.

Hold that against what a thirty-second broadcast spot normally costs in people, days and insurance, and the pull becomes obvious. This is not a story about executives who do not understand craft. It is a story about a cost structure that changed by an order of magnitude while the quality argument was still being had.

The first one ran in 2024. Coca-Cola described it in its own release as a global milestone, the first entirely generative-AI-created film on broadcast media, set to the music of the company's 1955 ad. The reception was not warm. The Verge called it an eyesore. Social media settled on AI slop.

A year later the company made another one.

A first attempt is a bet. A second attempt, after the first was publicly dismantled, is a position — and the people who ran it said so plainly rather than hiding behind the launch.

We need to keep moving forward and pushing the envelope. The genie is out of the bottle.

Pratik Thakar, Coca-Cola's head of generative AI, on the 2025 advertising, as reported November 2025

In the same reporting he addressed the previous year directly: people criticized the craftsmanship, and this version was ten times better. And then the sentence that actually contains the strategy: the company cannot keep everyone one hundred percent happy, and proceeds where most consumers respond well.

Read that as a trade rather than an opinion, because that is what it is. Coca-Cola weighed a production economics gain of roughly an order of magnitude against a loud, visible and repeated authenticity complaint, and took the gain. Twice.

Here is the part worth sitting with: on the evidence available, that may well be the right call for Coca-Cola. Nobody pays a premium for a Coke because a person drew the truck. The brand's price is held up by availability, memory and a hundred years of distribution. The authenticity complaint is real, it is loud, and it is landing on an asset the company was not charging for.

Which is exactly why the case does not transfer to the brand in your meeting, and exactly why quoting it as precedent is the mistake both sides are about to make.

When this argument reaches for evidence, it reaches for the trust literature, and the trust literature does not say what either side assumes.

The most directly relevant recent experiment — 617 participants, a two-by-two design, published in Behavioral Sciences in 2026 — tested what happens when content carries an AI disclosure label. The label measurably reduced perceived authenticity. It had no direct effect on trust or on intention to act at all.

That finding is bad news for both people in the room.

It is bad news for the objection, because the straightforward version of it — people will not trust us — is not what was measured. If you walk in saying AI will cost you trust, the first person to read the paper wins the meeting back.

It is bad news for the enthusiasm, because the damage did not disappear; it moved. In the same model, lower perceived authenticity predicted higher perceived risk, and higher perceived risk predicted lower trust and lower adoption. The effect is real and it is indirect. It runs through authenticity, which means it shows up in the places authenticity is doing commercial work — and for a premium brand, the place authenticity is doing commercial work is the price.

The wider literature points the same way without settling it. A 2025 systematic review covering 97 peer-reviewed articles on virtual influencers finds trust is built by anthropomorphic traits, content quality and identity cohesion, and undermined by the uncanny valley, perceived inauthenticity, and this: virtual influencers are generally perceived as carrying greater social-psychological distance than human ones. That distance costs a mass brand very little. It is close to the entire proposition of a premium one.

Registered before writing, because the thin parts matter more than usual here.

Labels are not influencers. The 617-person experiment tested disclosure labels on content, not a brand replacing a human spokesperson with a synthetic one. The mechanism is likely to carry. The size of the effect is not established for the harder case.

The review reports no effect sizes. The 97-article synthesis is directional. It tells you which way the forces point and not how hard they push, which is precisely the number this decision wants.

Backlash is not sales. Coca-Cola's two rounds of criticism are visible, measurable and almost entirely on social media. What happened to volume is not public. A company that runs the same play twice has seen something the audience has not, and reading loud complaints as commercial damage is the error that makes marketers sound unserious in finance meetings.

The two most on-point papers are behind paywalls. A 2024 study in the Journal of Consumer Behaviour on virtual influencer disclosure and brand trust, and a 2025 study in Psychology & Marketing on virtual influencers and luxury perception specifically, both sit directly on this question and neither is quoted here, because I could not read them. If this decision is worth real money to you, those two are worth the access fee.

Coca-Cola did not win an argument about authenticity. It declined to have one, because it had already worked out that authenticity was not the line item under threat.

The brand in your meeting may not have that luxury, and the difference between the two situations is a number nobody in the room is holding. Until somebody measures it, the loudest person wins — and the loudest person is usually the one with the cost line.

Related from Sound Decisions: A Third of the Shelf, Two Percent of the Growth · Anyone Can Look Like You Now · The Memory Isn't Yours

Chase Cut 395,000 Websites. Nine years later the industry caught up.

The waste in programmatic advertising was measured, named, costed and publicly pledged against. Three months after the pledges, sixteen of the seventeen companies that had measured it were still buying it. What finally moved the number was not better data.

You know the media review. The deck opens on reach, because reach is the number everyone in the room agrees on, and it closes on a recommendation to shift some budget from one platform to another. Somewhere in the middle there is a slide about brand safety with a logo on it. Nobody asks how many websites the money is actually landing on, because that number is not on any slide, and the one person who could produce it works at the agency.

If you did ask, and the answer came back in five figures, you would not know whether that was good or bad. That is the honest position most marketing leaders are in, and it is not a competence problem. It is that the instrument that would tell you has only recently existed.

Read that twice, because both halves are true. It is a record and it is less than half. The best quarter the industry has ever measured is one in which fifty-five cents of the programmatic dollar did not become a qualified impression.

The comparison that matters is where it started. In December 2023, the Association of National Advertisers published a supply-chain study with twenty-one major marketers and found that about thirty-six cents of every programmatic dollar reached the consumer. Twenty-nine percent went to ad-tech transaction costs. Thirty-five percent went to low-value environments — chiefly made-for-advertising sites, which had grown from roughly five percent of web auctions in 2020 to nearly thirty percent by the middle of 2023. Campaigns in the study were running on an average of 44,000 websites.

So the number has moved, genuinely, from about 36 to 45.1 over two and a half years. Anyone telling you nothing has improved is wrong. Anyone telling you it is fixed is also wrong, and the gap between those two claims is where most media budgets live.

Here is the sequence worth holding in order, because the order is the argument.

December 2023: twenty-one of the largest advertisers in the world open their programmatic supply chains, measure the waste, and publish it. The study puts a number on the opportunity — around twenty-two billion dollars, a quarter of the open-web market it examined. The industry response is immediate and loud. Pledges follow.

March 2024: an independent analysis finds that sixteen of the seventeen marketers who had been disclosed in that study still had impressions running on made-for-advertising sites. Ads from major packaged-goods companies were still landing on clickbait inventory. One campaign in the analysis reached a single person at an effective cost of $5,491 per thousand impressions.

Three months. Full visibility. Public commitment. Sixteen of seventeen.

That is the finding this piece exists for, and it is not a story about incompetence. These are sophisticated organizations with real analysts and real money at stake. They measured the problem precisely and then kept doing it, which tells you the constraint was never information.

Programmatic has entered an accountability era. Transparency and efficiency are now table stakes. What separates winners is disciplined execution.

Bob Liodice, CEO, Association of National Advertisers, on the Q4 2025 benchmark, February 2026

He is right, and the sentence is doing more work than it looks. If transparency is table stakes, transparency is not the differentiator. Everyone can see it now. What separates the top from the bottom has to be something other than knowing.

The Q2 2026 benchmark answers this directly, and the answer is almost comically unglamorous.

Higher-performing advertisers ran their campaigns across roughly 19,000 fewer domains and apps than lower-performing ones. Same market, same tools, same visibility. The ones getting more of their money into real impressions were the ones buying in fewer places.

The spread that produces is not marginal. Higher performers converted 52.3 percent of spend into qualified impressions against 31.1 percent for lower performers — a gap of twenty-one points. Looked at as price, lower performers were paying a nominal $6.55 CPM for media whose true cost per qualified impression was $13.80. They were buying at half price and paying double.

And one named case in the same reporting: Kimberly-Clark cut its true cost per qualified impression by 31.4 percent and raised qualified impressions per dollar by 45.7 percent.

Put the two together and the shape is clear. In 2017 one advertiser cut 98.75 percent of its placements on judgment, with no benchmark to point at and no proof beyond a few days of flat metrics. In 2026 the industry built the instrument, ran it quarterly, and found that the advertisers at the top of the table are the ones doing what Chase did.

Nine years of measurement to arrive at subtraction.

Because subtraction has no owner. Every other move in media has a department behind it, a vendor who benefits, a slide that makes someone look busy. A shorter list makes the plan smaller, the reporting thinner and the agency's job easier to question. There is no one in the room whose incentive is to argue for buying in fewer places.

And because the measurement, for most of that period, could not see the difference. If your reporting credits the last click, a placement that caught someone who was already buying looks identical to a placement that caused something. The waste was invisible at exactly the point where a decision would have been made about it, which is why visibility arrived years before behavior changed — and why sixteen of seventeen marketers could see the problem clearly and keep funding it.

Registered before writing.

The direction of causation is not established. Higher performers run fewer domains. Whether cutting domains makes you a higher performer, or being a higher-performing organization causes you to do both, is not something a benchmark can settle. The honest claim is that the association is strong and consistent, not that the lever is proven.

The benchmark measures delivery, not outcomes. "Qualified impression" means fraud-free, viewable, measurable and not MFA. That is a floor, not a result. An impression can clear every one of those tests and still sell nothing.

Chase is one company, briefly observed. See the case above. It is the strongest anecdote in the field and it is still an anecdote.

These figures are programmatic, and programmatic is now not the whole picture. Connected TV was around forty percent of the tracked spend by late 2025, and private marketplaces over ninety percent of median spend — a very different buying structure from the open-web auctions the 2023 study examined. The composition shifted while the argument was being had.

The thing that makes this decision hard is that it looks like doing less. A shorter list is a smaller plan, and smaller plans are harder to defend in a room that has spent a decade being rewarded for scale.

But the record quarter in the industry's own data is 45.1 percent, and the advertisers at the top of it got there by refusing most of what was offered. Chase worked that out in a week in 2017, without a benchmark, and took the hit of looking like it was retreating. The number everyone else needed took nine more years to build.

Related from Sound Decisions: Most of Your Budget Is Rent · The Budget Line · Nine Forecasts and No Evidence

What 1,800 Comments About Marvel’s Wolverine Actually Said

Across six comment sections and roughly 1,800 comments, the audience's argument about Marvel's Wolverine replaced itself at every stage of the launch. One word survived all three.

On September 3, 2026, PlayStation posted the launch trailer for Marvel's Wolverine. Twelve days to release, five years since the reveal, a character with forty years of audience behind him. The comments filled up the way they always do.

If you ran that launch, you would have watched the numbers climb and assumed you were watching people argue about Wolverine. You were not. The single most common thing people said under that trailer, and under the gameplay trailer before it, was about whether the game would be sold on a disc.

Not price. Not graphics. Not the story, the length, or the reported budget. Plastic.

Six rooms, one launch

I read six comment sections around the launch of Marvel's Wolverine: two of PlayStation's own trailers, two independent reviews published at embargo lift, one large reaction video, and one post-mortem published two weeks after release. About 1,800 top-level comments in total, read on September 29, 2026.

The method matters, because it is what makes the result trustworthy rather than anecdotal. Themes were not chosen by me in advance and then hunted for. The first pass reported the most frequent words and phrases in each section, counted once per comment, so a number means how many people said a thing rather than how often it was typed. Only then were the themes named, from what the audience had actually used, and counted across every section.

That sequence matters more than it sounds. My first attempt did it the other way round — five sensible-looking themes, named in advance: price, length, open world, voice acting, PC release. The biggest of them appeared in five comments out of four hundred. Every one of those five guesses was wrong, and a pre-registered instrument was the only reason I found out rather than writing them up.

The argument before the game existed

Under the marketing, two things were being discussed and nothing else came close. Discs, at 13.8%. And physical copies as a category, at 11.8% — the same conversation arriving in different words, with the phrase "physical disc" appearing on its own in 3.5% of comments.

The only other theme with real presence was blood, at 4.3%. That is the promise the marketing made — a violent, adult Wolverine after years of Spider-Man — and it landed. People came to the trailers to check that the violence was real and to ask whether they could buy a copy they would still own in ten years.

Neither of those is a question the marketing was answering. One is a merchandising decision. The other is a format decision. Both were being raised, at volume, in a channel owned by the people who could have answered them.

Then it all changes

By the time people had played the game, the disc conversation was gone. Not diminished — gone. Zero mentions in two hundred comments.

In its place: length. Eight per cent of comments two weeks after release mention an hour, most of them reacting to how quickly the game ends. The most-liked version of it is four words long and mostly punctuation. Alongside it, the story, at 4%, and the characters by name — Logan, Jean, Marcus — in double digits.

So the launch produced three separate conversations. Before release, an ownership argument. At review, no dominant argument at all: the top theme in those sections reaches only 5%, and the vocabulary is diffuse enough that the sections are barely readable as a conversation about the game. After play, a value argument about length and story.

Almost no vocabulary crosses between them. The people are presumably overlapping. The argument is not.

"So this is what a 300 million dollar game looks like."

Comment under a PlayStation trailer, one of one in four hundred

The number I went looking for, and did not find

Marvel's Wolverine reportedly cost $305 million to make — a figure that exists publicly only because Insomniac was breached in December 2023 and the documents leaked. Sony has never confirmed it. The same leak indicated the game needed roughly six million copies at full price to break even. Analysts estimated 1.9 million sold in its first three days.

That is a good story, and I expected the audience to be telling it. In one section they were: under a reaction video with 3.9 million views, 5% of comments say "300 million" and 3% say "budget". People there are pricing the game against a number its publisher never gave them, which is a genuinely strange thing for a market to do.

But across the other three sections, the same phrase appears in one comment out of four hundred, twice, and in none at all two weeks after release. The budget argument is real, loud, and confined to a single room. Had I read that one section, or scrolled the top comments anywhere, I would have written a confident piece about a market that judges games by their leaked budgets. The counting is what stopped me.

What survives

One word appears in every section: Insomniac. It runs at 4.8% under the marketing, 4.8% at review, 6% under the reaction, and 7% two weeks after release.

It is the only theme that rises as the reception falls.

That is the finding a brand should care about most, and it is not about Wolverine. The game absorbed the review scores. The studio absorbed the reputation. For anyone running a studio, a label, a production house or any brand that makes things under its own name for someone else's franchise, that is an asset depreciating on a product you may not control.

What the evidence cannot say

These are public voices, not customers: people who chose to comment under a video, which is not the same as people who bought the game. Two hundred top-level comments per video, in YouTube's relevance order — a sample, not a census, and replies are not included.

Counting is literal. A comment mentioning blood may be objecting to it. Themes are counted, not judged, and nothing here measures sentiment.

Two of the sections I read could not carry a finding at all, and I would rather say so than quietly drop them: one was 45% dominated by a running joke, another 28% about the creator rather than the game. Both are in the study and neither is in the table.

The budget, break-even and sales figures are leaked or estimated and have never been confirmed by Sony or Insomniac. They are reported here as what is publicly claimed, not as fact.

Built on the ADMADS Engine: sourced, dated evidence, with its limits stated. How these pieces are evidenced.

What the Ad Reads Bought

Two companies bought the same kind of attention on YouTube. One grew while spending a smaller share each year. The other spent more than half its revenue on advertising and lost customers.

If you run a consumer business, someone has told you to buy creator ad reads, and someone else has told you they are a waste of money. A popular video by the YouTuber Micro recently made the second case well: the sponsors you hear on YouTube are often young companies spending investors' money to buy attention fast, and the audience carries the risk.

I wanted to know which side the evidence is on. The answer turned out to be neither, and the reason is the useful part.

The test, written down first

I took six brands known for heavy YouTube sponsorship and paired each with a rival in the same category that mostly buys its attention elsewhere: Raycon and Bose, BetterHelp and Talkspace, Manscaped and Gillette, Squarespace and GoDaddy, NordVPN and Norton, and Raid: Shadow Legends and Candy Crush.

Before pulling any data, I wrote down the test. For each brand, how much of its best year of search demand does it still have? If the sponsors kept at least 20 points less than their rivals in four or more of the six pairs, the claim that ad reads don't build lasting demand would hold. At two or fewer, it would fail. In between, the answer would be that the data can't say.

Then I measured it three ways. Each one is a reasonable way to count search interest. They differ in the source and the start date.

The verdict depends on where you look. On Google Trends, NordVPN is still at its peak. On Google Ads data over the same months, it is down 43 percent. Three of the six pairs change sides from one view to the next, and on the strictest view the claim fails outright.

Only one brand told the same story every time. Squarespace kept its demand.

So search data can't tell a founder whether ad reads work. It can tell you that anyone who sounds certain about it probably picked one source and one start date, and stopped there.

The money

Search interest is attention. A business runs on customers, and customers show up in the accounts. So I went to the annual filings, where there were any. Four of the six sponsors are private or file outside the United States. Two are worth reading closely, because their filings say what they spent and what it bought.

BetterHelp, owned by Teladoc Health, raised its advertising and marketing 3 percent in 2024 while its revenue and its customers both fell. More than half of every dollar it earned went back into advertising. In 2025 its revenue fell another 9 percent, to $950.4 million, and the segment's adjusted earnings fell 46 percent. Teladoc's filing says plainly that reaching paying users through advertising channels remains critical to its success.

Squarespace went the other way. Between 2021 and 2023, its last full year as a public company, revenue grew from $784 million to $1,012 million. Its advertising went from about 35 percent of revenue to about 25 percent. It spent fewer dollars on advertising in 2023 than in 2021, and still grew by more than a quarter.

GoDaddy spends about 5.5 percent of revenue on advertising. Norton's parent, Gen Digital, spends about 12 percent, though its latest growth includes an acquisition, so it says little on its own.

The rival that changed who pays

The most useful comparison is BetterHelp against Talkspace, because they sell almost exactly the same thing: therapy on your phone. Talkspace spent about 13 percent of its 2025 revenue on advertising, a fifth less in dollars than in 2022, while its revenue nearly doubled over the same years.

It didn't get there by out-advertising BetterHelp. It changed who pays. In 2025, Talkspace's revenue from health plans paying for their members' sessions grew 38 percent, to $171.5 million, three quarters of the company. Its own cash-pay business, the one that looks most like BetterHelp's, shrank 30 percent on purpose; its annual report calls that an intentional decision to focus marketing on insured members. Net income rose from $1.1 million to $7.8 million.

That route has a price. Insurers pay less per session than someone paying full price, and Talkspace's gross margin slipped as the mix shifted. It also takes contracts with each health plan, therapists who can bill them, and patience. None of that can be bought in a week. BetterHelp began accepting insurance in 2025, and by mid-2026 Teladoc was telling investors that cash-pay revenue was falling faster than insurance revenue could replace it.

The full BetterHelp and Talkspace story is on Sound Decisions.

Your call: will BetterHelp's 2026 revenue beat, meet or miss Teladoc's own lowered forecast? Make your prediction. Voting closes October 21.

What separates them

Squarespace, BetterHelp and Talkspace all bought attention. What they did with it was different, and it comes down to two routes.

My reading, and it is a reading rather than a measurement: either the customers you buy stay, or someone else brings them to you. Squarespace looks like the first. A website is hard to move once it is built, so each customer bought keeps paying, advertising compounds, and the share of revenue it needs can fall. Talkspace is the second. A health plan with its own reason to send members does much of the finding, so the ad budget can shrink while the business grows.

BetterHelp, in 2024, had neither working. A therapy subscription is easy to pause, and nearly every customer was one it paid to find. When that happens, each year starts close to zero and the bill rises just to stand still.

The filings show the spending and the outcomes. They do not prove the cause. None of these companies publishes retention in a form that would settle it, and I would want to see that before saying more.

For scale, creator ad reads are priced at roughly $10 to $50 per thousand views, depending on the niche, according to 2026 creator rate surveys. That is an estimate, and the price is not really the question. What each dollar brings back is.

Related from Sound Decisions: Most of Your Budget Is Rent · The Flip Flop That Isn't

Built on the ADMADS Engine: sourced, dated evidence, with its limits stated. How these pieces are evidenced.

Coke Built It. Pepsi Bought It.

A jab step costs the attacker a step and the defender his balance. Two prebiotic soda startups raised under $100 million between them and made Coca-Cola and PepsiCo commit. One built. One bought for $1.95 billion. The third answer belonged to the challenger.

Carmelo Anthony was the NBA's scoring champion in 2013, averaging 28.7 points a game, and he still holds the New York Knicks' single-game record of 62. He was known for strength, quickness and a variety of crafty moves close to the basket. One move built for exactly that spot is the jab step: a short, sharp step toward the defender that looks like the first step of a drive.

The move costs the attacker almost nothing. He has not committed his weight. The defender has no such luxury. People take about 190 milliseconds on average just to react to something they see, and the time grows as the number of things that might happen grows. A player facing someone who can shoot, drive left or drive right has to prepare for all three, and a convincing fake forces him to commit his weight to one before he knows which is real. The attacker spends a step. The defender spends his balance.

Challenger brands do this to the largest companies in their category. Prebiotic soda is the clearest recent case, because the incumbents' answers are on the record.

A small bet that forced a large one

Poppi began in Houston as Mother Beverage, an apple cider vinegar drink. It took $400,000 on Shark Tank in 2018, rebranded as Poppi in 2020, and raised a $13.5 million Series A in 2021 and a $25 million Series B in 2022. Olipop, founded the same year, had raised $55.4 million in total by January 2023. Between them, well under $100 million of disclosed venture money.

By 2023 both were past $100 million in revenue. Olipop reported $100 million in gross revenue in the first half of that year alone, and its root beer outsold A&W's. Two small companies had opened a category on the soda shelf, in the space Coca-Cola and PepsiCo consider theirs. Each incumbent had to decide how much weight to commit.

Three answers to the same fake

PepsiCo committed its full weight. It paid forty to fifty times the challenger's venture funding and got speed: a brand with demand already built, and the leading position in the category on the day the deal closed. It also got the brand's history. In 2024 a California class action argued that one can held too little prebiotic fiber to deliver the gut benefit the marketing implied. The company agreed in July 2025 to settle for $8.9 million, without admitting wrongdoing.

Coca-Cola kept its weight back. Simply Pop cost a line extension under a name shoppers already trust for juice. It is cheap, and it can be quietly withdrawn if it fails. What it did not buy is a head start. Poppi and Olipop spent years building the pull that keeps a product on a shelf; Simply Pop started without it.

Olipop took the third option, one only the challenger had. It raised again at a valuation just below what PepsiCo paid for its rival, and kept both incumbents guessing.

Who was right is not answerable yet

Nobody outside the two companies knows. PepsiCo does not break out poppi's sales in its filings, and Coca-Cola does not break out Simply Pop's. The fair test is shelf position and share over 2026 and 2027, and I will check both answers against what the companies and retailers report by September 2027.

What can be said now is what each answer risks. A buy that overpays cannot be undone. A build that arrives without demand can sit on the shelf and be delisted without anyone noticing. Both are rational. They are bets on different things: PepsiCo that the category's demand is real and already owned by someone, Coca-Cola that a trusted name can catch up.

What this means if you are the challenger

The jab step works because the attacker keeps his options live and the defender cannot. A challenger's options are the same three the incumbents faced from the other side: keep growing and stay independent, be copied, or be bought. The founders who get to choose among them are the ones who built demand an incumbent can see in its own numbers, and whose claims hold up when a lawyer or a foreign regulator reads the can.

Most challengers never force the question. They are ignored rather than copied or bought, because nothing they do shows up in the incumbent's numbers. The fake only works if it looks like a real drive.

Related from Sound Decisions: Poppi Didn’t Go Viral by Accident · The Ground Nobody Was Defending · City Stopped Tuning and Bought a Finisher

City Stopped Tuning and Bought a Finisher

Manchester City won three league titles in four years and kept losing the one game that decided Europe. In 2022 it stopped tuning, paid £51.2 million for one striker, and changed the team around him. The rebuild is the part brand teams skip.

Most brand teams know this position. The business runs well. Distribution is solid, execution is disciplined, the plan gets delivered. And yet the result that decides the year keeps going the other way: the national reset, the anchor retailer's review, the one launch the board is watching. The ordinary weeks are won. The weeks that matter are not.

The usual answer is to tune. Tighten the plan, sharpen the execution, do what already works a little better. Manchester City spent two seasons doing exactly that, and then made a different call. The club's own record shows what it cost and what it bought.

The system that won almost everything

In 2020–21, City were confirmed as Premier League champions for the third time in four seasons. They spent the first months of that season without a recognized first-team striker: injuries to Sergio Agüero, carried over from the previous season, and then to Gabriel Jesus, left the squad without one until the beginning of December. They won the league anyway.

Then the season's biggest game went the other way. On May 29, 2021, City lost the Champions League final 1–0 to Chelsea.

That summer Agüero left as the club's all-time leading goalscorer. City pursued Harry Kane, then at Tottenham. On August 25, 2021, Kane announced he was staying. City did not sign a replacement striker. They kept the system and tuned it.

It worked, mostly. City retained the league title in 2021–22. And in the Champions League semi-final they lost again, 6–5 on aggregate to Real Madrid after extra time.

The call

In May 2022, City agreed to sign Erling Haaland from Borussia Dortmund for a fee of €60 million (£51.2 million). He was known for his positioning and finishing, especially inside the penalty area, and he was joining a club that had lost its all-time top scorer a year earlier and not replaced him.

The fee was the smaller part of the decision. City did not simply drop Haaland into the old team; its shape changed around him. John Stones, a center-back, spent the season defending in the back line when City did not have the ball and stepping into midfield beside Rodri when they did. The club added a player it had not had, and moved another to keep the midfield it was built on.

On June 3, 2023, City beat Manchester United 2–1 in the FA Cup final. On June 10, they beat Inter Milan in Istanbul to win the Champions League for the first time. With the league already won, it was only the second continental treble by an English men's team, after Manchester United in 1998–99.

What this is not evidence of

One season proves less than it seems to. Football has plenty of expensive strikers who did not fit, and City were already the strongest side in England; the treble is not proof that any team adding any finisher wins. Haaland also arrived with a record. He had been named Bundesliga Player of the Season at Dortmund. City did not buy a hope.

And a team built around one player carries a risk it did not carry before. If the finisher is injured or leaves, the shape built to serve him has to change again.

The brand version of the same call

Strip out the football and the decision is common. A brand is strong where it is strong, and it keeps losing in one specific place: the club channel it cannot crack, the premium tier its line doesn't reach, the occasion a challenger owns. The team's reflex is to tune. Better promotions, better execution, a sharper plan.

Sometimes the loss is not a tuning problem. It is a missing capability: a product, a partner, a hire or an acquisition that does the one job the current system cannot. City's two years show the sequence. First, the losses clustered in one kind of game. Second, tuning was tried and won everything except that game. Third, the fix was a specialist, and it only worked because the rest of the team was changed to make room for it.

The third step is the one brand teams skip. They add the capability and ask it to fit into the existing plan unchanged. A premium line sold through the same promotions calendar as the core range. A new channel run by the team that runs the old one, on the old one's targets. That is signing Haaland and asking him to drop deep.

Related from Sound Decisions: On Didn’t Sign Mbappe · The Ground Nobody Was Defending · Coke Built It. Pepsi Bought It.

AI Agents Are Learning to Buy. The Brands Haven’t Said a Word.

Google published an open standard for AI agents to buy from retailers. The card networks, platforms and biggest retailers have written agentic commerce into their SEC filings. Twelve of the largest packaged-goods makers haven’t used the word once. When an agent shops, the pack and the ad go unseen.

A shopper tells an AI assistant to reorder paper towels under $25. The assistant searches, compares, picks a brand and pays. Nobody walks an aisle, sees an end cap or scrolls past an ad. The industry has started calling this agentic commerce: AI agents that find, choose and pay for things on someone's behalf. It is early. The companies that sit between brands and shoppers are already writing it into their plans. The brands mostly aren't.

The decision that set the terms

In January 2026, Google announced the Universal Commerce Protocol, an open-source standard for how AI agents talk to retailers' systems. The choice is worth reading as a decision. Google could have built a closed checkout inside its own assistant. It published a shared standard instead, so any agent can transact with any retailer that adopts it.

The rest of the plumbing moved the same month. Commercetools launched a product that connects an existing commerce system to ChatGPT, Google Gemini and Microsoft Copilot without rebuilding it. In China, Alibaba's Qwen assistant already lets people complete purchases inside the chat, paid through Alipay's AI payment service.

Who has written it down

Public companies tell investors what they think will matter, in filings they can be held to. So the useful question is who has put "agentic commerce" in one, and who hasn't.

Between January 2024 and September 26, 2026, 141 annual, quarterly and current reports filed with the SEC used the phrase, from 41 companies. The first was the advertising holding company Interpublic, in April 2025. Then the money moved in: American Express, PayPal and Bread Financial in July 2025, Mastercard and Visa that autumn, Capital One and Synchrony in early 2026. The platforms followed: Shopify, Etsy, Klaviyo, Criteo. So did retailers: Ulta Beauty, Macy's, Kohl's, Revolve.

Why the silence matters to a brand

Every layer in that map sits between a brand and its shopper, and every one of them has told investors it is preparing. When a person shops, a brand can win with a pack, a price point, a shelf position or an ad. When an agent shops, it works from whatever the retailer and the protocol hand it: the product's attributes, price, ratings and availability, in a form a machine can read. The things brands spend most of their money on are the things an agent doesn't see.

The purchases agents suit first are the ones packaged goods depend on. Recurring buys, subscriptions and restocking are predictable and standardized, which is exactly where delegating to software is easiest. That describes a large part of a grocery basket.

What this is not evidence of

A filing is not a strategy. Brand owners may well be working on this without writing about it, and a phrase in a card lender's quarterly report can be one line in a list of risks. The counts show where the subject has reached the level of investor disclosure, not how much anyone is spending. Agentic commerce is also small today. Nobody outside the companies involved knows how much buying agents actually do.

What the filings do show is order. The standard, the payment networks, the platforms and the largest retailers moved first. The brands whose products get chosen are last, and so far silent.

Make a call: how many of those twelve brand owners will name agentic commerce in a filing by mid-2027? Cast your vote, free, settled by the same public search.

Related from Sound Decisions: Change the Question, Change the Ranking · Coke Built It. Pepsi Bought It. · Getting Into Walmart Is the Dream

The Flip Flop That Isn't

If Omnicom wins Coca-Cola's North American media, it won't have switched sides. It will be working both. And the filings show the two cola companies betting in opposite directions.

In one of our group chats this week, someone summed up the cola agency shuffle in one line: if Omnicom wins Coke, they basically flip flop.

It's a good line. It's also not quite what would happen, and the difference is the most useful part of the story.

What moved

In March 2025, Publicis took Coca-Cola's North American media from WPP. On September 2, 2026, PepsiCo moved its global media account, about $1.7 billion of 2025 spend by COMvergence's estimate, from Omnicom to Publicis. There was no formal pitch, and Omnicom had held the business in most markets for more than 25 years.

The same week, Publicis withdrew from the pitch for Coca-Cola's international media, and Coke kept that business with WPP. Coca-Cola's North American media, estimated at $700 to $800 million, is now up for review. Omnicom and Dentsu are understood to be the two groups competing for it. Coca-Cola hasn't confirmed the shortlist.

Why it isn't a flip flop

Omnicom lost PepsiCo's media, but it didn't lose PepsiCo. It still handles Pepsi's creative, sports and PR work, including in North America.

So if Omnicom wins, one holding company would plan and buy Coca-Cola's media in North America while making Pepsi's advertising in the same market. That isn't a switch of sides. It's a seat at both tables, divided by discipline.

Whether Coca-Cola accepts that division is the real question in this review, more than which agency has the better deck.

Two companies, opposite bets

Neither move was a one-year blip. Over two years, Coca-Cola's advertising is up 8.0 percent and PepsiCo's is down 10.5 percent.

Two cautions. These are company-level advertising lines, not media billings, so they don't match the account values the trade press reports. And they describe what each company spent, not why.

What they do show is two different bets. PepsiCo is spending less and consolidating what's left into one partner and one operating model, which Publicis will run across more than 200 markets. Coca-Cola is spending more, but in North America it's the one being reorganized by moves it didn't make.

What each side is really choosing

Coca-Cola went from running a planned global review to choosing between the two groups still available. It has said it wants to move from traditional media planning toward reaching people through technology, including agentic tools. It isn't alone: so far in 2026, 350 annual reports filed with the SEC mention "agentic," up from 72 in all of 2025. That counts filings, not companies, and a mention isn't a commitment, but it shows what public companies now think is worth telling investors.

Omnicom is trying to replace a top-three client within weeks of losing it, while still integrating Interpublic, which it acquired last fall. Winning Coke would also test whether clients will accept one group on both sides of the cola war, split by discipline.

Dentsu already works with Coca-Cola in Japan and Korea. This is a chance to widen a relationship it already has.

PepsiCo keeps Omnicom on its creative. If Omnicom wins, Pepsi's creative partner would also be buying media for its biggest rival. Pepsi has said nothing publicly about that possibility.

No prediction market has priced the outcome. As of September 25, 2026, neither Polymarket nor Kalshi lists a market on it. Make your call at cultmath.com/predict/coke.

What it means if you advise brands

Your agency's other clients are part of your risk. Coca-Cola didn't lose leverage because of anything it did. It lost leverage because its agency took the rival.

Long relationships can end without a pitch. Twenty-five years of media work ended with no review. Decisions this size increasingly sit with transformation programs and the executive team, not only with the marketing seat.

Name what you're buying. Coca-Cola stated its criterion in public. A brief that names a capability gets a real answer. A brief that asks for "strategic partnership" gets a longer deck.

Related from Sound Decisions: Takis Beats Doritos · Same Cocoa Shock, Two Answers

Built on the ADMADS Engine: sourced, dated evidence, with its limits stated. How these pieces are evidenced.

Takis Beats Doritos

In US searches, one Grupo Bimbo brand draws nearly twice the attention of Doritos. Measure a big food portfolio the same way and its strongest brands turn out to sit where attention is thinnest.

You know the portfolio review. The brands are ranked on a slide by revenue, then by margin, then by share of shelf. The ones at the top got the most money last year and will ask for more this year, and the case is easy to make, because every number in the deck says they are winning. The question almost nobody in the room asks is whether anyone is still looking for them.

That question is harder, because the deck cannot answer it. Sales record what people bought after they had already decided. Search sits closer to the moment of deciding, when someone is curious enough about a brand to type its name. If you are the one deciding which brands to fund, it is the difference between reading last year and reading next year.

General Mills gave a clean example of what is at stake in June 2025, when it completed the sale of its US yogurt business, Yoplait included, to Lactalis. That business still brought in about $1.2 billion of sales in its final year. On a revenue slide, it was a large business.

So I ran a simple test on nine US grocery categories to see what the same portfolio looks like measured by attention instead. For each category I named a set of three to six brands, pulled twelve months of Google search volume for each name, and worked out each brand's share of the searches in that set. The whole dataset cost less than a dollar.

The aisle with the most attention

The largest set by a distance is salty snacks, at just over a million searches a month. PepsiCo owns four of its six names: Doritos, Cheetos, Lay's and Ruffles. The single most-searched brand is none of them. It is Takis, made by Barcel, part of Mexico's Grupo Bimbo. Takis draws 36.3 percent of the searches in the set. Doritos draws 19.8. Lay's draws 10.8, level with Pringles.

To be fair to PepsiCo, its four brands together hold 52.8 percent, well ahead of Takis. The company is not losing the aisle. But no single one of its brands draws the attention that one rolled tortilla chip does, and a portfolio review built on sales would never show you that.

Nine categories, ranked by attention

Here are all nine sets, ranked by how many searches the named brands draw between them each month.

Read the table from the top. The three largest sets, salty snacks, confectionery and better-for-you drinks, have no General Mills brand in them. General Mills leads twice: Cheerios in cereal at 31.1 percent, the fourth-largest set, and Betty Crocker in baking mixes at 45.8 percent, the seventh. Totino's and Progresso hold second place in theirs. Nature Valley holds 14.6 percent of bar searches, under a third of Clif Bar's 48.9, and Clif Bar belongs to Mondelez.

Where General Mills leads, the pool of attention is modest. Where the pool is largest, it isn't there.

A decision that has already been made once

Now go back to yogurt. The business General Mills sold was large by revenue. By attention, it had already become a small brand: Yoplait draws 7.3 percent of the searches in its set, fifth of five, while Chobani draws 54.7.

The sale was made on financial numbers, as it should have been. The point is that the attention numbers were pointing the same way, and they cost cents to pull. Nature Valley sits in a similar position today, a distant third in a smaller set. The same question applies to it, and it can be asked now, rather than after the sales line confirms the answer.

What a portfolio review usually measures instead

Revenue, share of shelf, margin by segment. None of them tells you whether anyone is looking for you. A brand can hold its shelf for years after attention has moved on, because shelves follow sales, sales follow habit, and habit is slow to break. That lag is exactly why attention is worth watching.

Research on share of search, starting with Les Binet's work for the IPA, found that a brand's share of searches tends to move ahead of its market share in the categories studied, often by several months. It is an early reading, not a verdict.

What this does not show

Search is not sales. Habit categories like cereal generate fewer searches per dollar than snacks people discover, share and talk about. A small pool of attention is not a small business: cereal remains a very large one. What the table shows is where new attention is going, not where the money is today.

Every share belongs to a named set. The brand lists are printed under the table so anyone can challenge them. Swap one brand and every number in that row changes.

Brand names are messy. Totino's is searched under two names, and the two are summed here. "Lay's" and "Lays" return the same combined figure, so it is counted once. Pillsbury is left out of General Mills' baking count, because General Mills sells Pillsbury dough while Pillsbury baking mixes are made by another company under license, and search can't tell the two apart.

This is not advice to buy a snack company. Salty snacks and confectionery have their own incumbents and their own economics. The piece names the question. It does not answer it.

If your strongest brands sit in your smallest pools of attention, the portfolio question is not which brand to grow. It is whether you are competing where people are actually looking, and what it would cost to get there.

Related from Sound Decisions: Same Cocoa Shock, Two Answers · Distribution Doesn't Create Demand · A Third of the Shelf

Built on the ADMADS Engine: sourced, dated evidence, with its limits stated. How these pieces are evidenced.

Same Cocoa Shock, Two Answers

Cocoa prices more than tripled, and Hershey and Mondelez each lost more than ten points of gross margin. One cut advertising by a fifth. The other kept spending. Both decisions are in the public filings.

If you run a brand, your budget arrives as a number from finance. You can argue about how to spend it. You rarely get to argue about the number itself, because it was set further up, by the margin the business expects to earn. When that margin moves, the budget moves with it, and the reason can sit a long way from anything marketing did.

The last two years produced an unusually clean example. I pulled the filed gross margins of seven US packaged food companies for their latest fiscal year, and what each spent on advertising, straight from their annual reports to the SEC.

Every margin fell. Five fell by less than two and a half points. Two fell by more than ten: Hershey, from 47.3 percent to 33.5, and Mondelez, from 39.1 percent to 28.4. They are the two largest chocolate businesses in the set. A year earlier, Hershey was earning more than a dozen points more on each dollar of sales than most of the companies around it. It now earns about the same.

The cause is on the commodity chart

Cocoa averaged about $2,340 a tonne in 2019. In 2024 it averaged $7,391. In January 2025 it reached $10,710, and it stayed above $8,000 through the first half of that year before falling back.

Both companies say so plainly. The more interesting question is why Hershey's hit landed in 2025 rather than 2024, when prices first spiked. The answer is hedging. In its 2024 annual report, Hershey said gains on the derivatives it uses to lock in future purchases more than offset higher commodity costs that year. In 2025 those gains turned into losses, and the company put the fall in margin largely down to higher commodity costs and unfavorable mark-to-market on those contracts.

That matters for how you read the table. Filed margins include these hedge swings, so the drop looks steeper than the change in the underlying business. Hershey's own guidance for its adjusted margin, which strips the swings out, was a decline of about 675 to 700 basis points for 2025. Roughly half the headline. The direction is the same either way.

Two answers

Here is where the two companies part. Mondelez cut advertising by 21.6 percent, to $1.66 billion. It did not hide the choice: its full-year results named lower advertising and consumer promotion costs among the things that lifted its adjusted operating margin. Hershey went the other way, up 2.0 percent across the year to $612 million, though not in a straight line. Its advertising rose 35.5 percent in the second quarter of 2025 and fell 5 percent in the third, mainly from lower agency fees.

Same shock. Same year. Opposite calls on the one budget line a brand team can see and argue for.

Who was right is not answerable yet

The honest answer is that nobody knows. Advertising works over years, and the real test is volume and share across 2026 and 2027. Cutting may have protected Mondelez's earnings at the cost of brand strength it is now paying to rebuild. Holding may have cost Hershey profit it never gets back. The filings will settle it in two years, and I will score this read against them when they land.

Not every cut is a margin decision

One more company in the table is worth a second look. Simply Good Foods, the smallest of the seven, cut advertising by more than either chocolate maker: down 26.1 percent. It was exposed to cocoa too. Its management said it had contracted cocoa at historically high prices to secure supply, and that this weighed heavily on margins in the second half of its fiscal year.

But its cut had a different cause. The company's results put the fall in marketing spend down to a planned reduction for Atkins, a brand losing distribution. From the outside, the Simply Good cut and the Mondelez cut look alike. Only the note in the filing tells you that one is a margin decision and the other is a brand decision. Anyone reading competitors' spend from the outside needs to read that far.

What this means for the marketing team

Budgets are set by margin, not by how well the marketing is working. A brand team with good results can lose a fifth of its spend because of a bad harvest thousands of miles away. Knowing which input costs sit under your category is therefore a marketing skill, not a procurement one. It tells you months ahead when the budget conversation is going to change, and it lets you walk into that conversation with a plan instead of a defense.

How this was measured

Gross margin is net sales minus cost of sales, divided by net sales, from each company's own filed figures for the same fiscal year. Kraft Heinz reports sales including certain taxes collected from customers, which slightly understates its margin next to the others. Companies define cost of sales differently, so compare the direction of change across companies, not the exact levels. Fiscal years differ: General Mills and Conagra end in May, Campbell's and Simply Good Foods in August, the rest in December, so each caught a different stretch of the price spike. Advertising is each company's own reported line and is not defined identically across them. Campbell's and Simply Good Foods both made acquisitions in 2024 (Sovos Brands and OWYN), which affects their year-on-year comparisons.

Related from Sound Decisions: Takis Beats Doritos · The Spend You Inherited Is Two Different Problems · A Third of the Shelf

Built on the ADMADS Engine: sourced, dated evidence, with its limits stated. How these pieces are evidenced.

How These Pieces Are Evidenced

Every figure in Sound Decisions carries a source and a date. This page says where those figures come from, what gets checked before a piece runs, and what the evidence cannot say. If you want to argue with a number here, this is the page that tells you how.

Where the figures come from

Company figures come from annual reports filed with the SEC, read from the filings themselves rather than from a summary. Each one carries the accession number of the filing it came from, so you can open the same document I did.

Market and category figures come from public statistical sources: the US Census Bureau for how many firms operate in a category and what they pay in wages; the Bureau of Labor Statistics and the Federal Reserve for prices, both what shoppers pay and what producers charge; the IMF for global commodity prices.

Search figures come from Google's own advertising data, which reports how many people search for a term in a month and how that has moved over a year. Attention figures come from Wikipedia pageviews, which measure curiosity rather than demand and are labeled that way wherever they appear.

Where a piece quotes what a company says about itself, it uses the company's own words from a filing, an earnings release or prepared remarks — not a paraphrase of them.

What gets checked before a piece runs

A control. No piece claims a whole market moved without measuring something outside that market. If a category's search demand fell by half, I measure eight to ten unrelated terms as well. If those fell the same way, the finding is a measurement problem and the piece does not run. If they held while the subject fell, the finding is real, and the control goes in the article where you can check it.

Whose view a measurement reports. A site-speed score from a testing service is one cold fetch from a data center, not what a visitor on a fast connection sees. Both readings can be true at once. Any measurement of that kind says which one it is.

What a market actually is. A company's market is where its customers are, not where its office is. Sizing a local market for a business that sells nationally is a mistake I have made and corrected, and every geography figure now has to state how it was established.

Absence is not zero. When a statistical source suppresses a figure or reports nothing, that is unknown, not zero. Totals say how many inputs were left out.

What the evidence cannot say

Public data has hard limits, and a piece that hides them is worth less than one that names them.

Private companies file nothing. Whole categories are visible only through their listed competitors. Where a private company matters to a story, the piece says what cannot be known about it.

Search volume is not demand. It is what people type into a search box. Google rounds and bands the figures, and the most recent month can be incomplete. Share of search is a share of a named set of brands: adding or dropping one moves every number, so the set is always listed.

Attention is not sales. Encyclopedia pageviews and comment counts measure curiosity and conversation. They are useful, they are weak, and they are never presented as evidence of what customers bought.

Public comment is not customer research. People commenting on a video are strangers in a category, not a company's customers. That distinction is kept in every piece that uses them.

Correlation stays correlation. Where a piece offers an explanation for a pattern — why two companies cut spending, why a kind of search disappeared — the explanation is labeled as a hypothesis, with what would confirm it.

Why the figures can be re-checked

Every gathering run is stored with its date, so a claim made in one piece can be measured again later. That is the point: a figure published in September can be re-run in March, and the difference is either a finding or a correction. Both are worth publishing.

It also means these pieces can be wrong in public. When a figure here turns out to be mistaken, the correction goes in the article, dated, rather than quietly replacing what was there.

Who built this

The gathering runs on the ADMADS Engine, a tool I built and use for client work. It pulls from public sources, stamps every value with its origin and pull time, and states what a given set of evidence cannot support. It does not write conclusions. The arguments in these pieces are mine, and so are the mistakes.

Built on the ADMADS Engine: sourced, dated evidence, with its limits stated.

Sound Decisions is analysis for general information. It is not investment, financial or legal advice, not a claim of wrongdoing by any company or person, and makes no prediction about any named company. Where the author has worked with a company named in a piece, the piece says so and uses only public information. © 2026 CULT+MATH LLC.

Your Next Client Is Already Inside This One. The pitch is the most expensive door.

Every professional firm, from ad agencies to law firms, finds most of its new work the same way: someone who already worked with them tells someone else. The decision that matters is where you spend the hours that produce that.

When I worked on the client side of a large packaged-food company, agency people regularly asked me for a way in. A meeting, an introduction, a name. I had a senior title and a narrow slice of budget authority, and I could not have handed anyone an account if I had wanted to. What stayed with me was how much effort went into getting through that door, and how little seemed to go into the accounts those same agencies already held.

That turned into a question I could test. Do agencies struggle to win work because they chase it, while the big strategy consultancies have it come to them? I wrote the hypothesis down before reading anything, and it did not survive. The channel is the same for every kind of professional firm. What differs is where each one spends its effort.

Ask agency owners where new business comes from and the answer barely moves. In SparkToro's 2024 survey of digital agencies, 66 percent said their top source was referrals from existing and past clients. Law firms report the same thing. Clio's 2025 report on solo and small firms found referrals were their strongest source of new clients, and BTI Consulting, which interviews corporate legal buyers every year, reports that up to 64 percent of them hire the first firm recommended to them.

There is no hidden channel the consultancies use and agencies miss. There is the person who already worked with you, and what they say about it afterward.

The door most agencies know best is the pitch. In 2023 the ANA and the 4A's, the trade bodies for advertisers and for agencies, published the first study to price a review from both sides of the table, from 329 interviews with marketers, procurement staff and agency people.

The rest of the report explains why that money is hard to earn back. Two in three clients said they kept their incumbent agency after their most recent review. Cost and price was the top factor in choosing a winner, named by 62 percent of clients, well ahead of creative execution at 45 percent. And the average agency in the study took part in 23 reviews over two years, about one a month.

Put together, the pitch looks like this: a large cost paid up front, for a minority chance at an account that is decided mostly on price. Some pitches are worth it. As a system for growth, it is the most expensive door in the building.

“We have one of the rarest agency partnerships in the history of our industry.”

Kate Trumbull, EVP and chief marketing officer, Domino's, April 2025

The strongest evidence on how firms grow inside accounts comes from law. Heidi Gardner at Harvard Law School's Center on the Legal Profession studied a decade of law-firm records. Partners who routinely brought other partners into their client work grew their own books of business significantly in the years that followed, even after controlling for how large their book was to begin with. The fear that big clients would use their weight to demand discounts turned out to be real, she found, but clients served by several practices were still more profitable on average over the long run.

For a small firm the translation is direct. Bringing a trusted specialist into a client's problem feels like giving margin away. The data says it is how the book grows.

If referral is the channel, the question becomes what produces a referral. Good work is required, and on its own it is not what clients talk about. A 2019 study of business clients of financial consultancies, published in Industrial Marketing Management, found that the experience of being served had a stronger effect on satisfaction, loyalty and word of mouth than the measured quality of the service. BTI, which ranks law firms using only their clients' ratings, ties four behaviors to higher profits, rate premiums and retention: commitment to help, focus on the client, understanding the client's business, and value for the fee.

Firms and clients also disagree about why relationships end. Setup, a firm that runs agency searches, surveys both sides every year. In its data, clients ranked problems with delivery as their second most common reason for leaving an agency. Agencies ranked it ninth. The ANA's 2023 report adds a second signal: half of clients named rising agency staff turnover as a factor in how often they call a review.

And clients who aren't pushed into reviews tend to stay. The ANA and the 4A's reported in April 2025 that agency-of-record relationships now average about seven years, and 8.1 years where the client has no mandatory review period, against as little as 3.8 years where reviews are frequent.

Nearly all of it is self-reported. Surveys reach firms that are still in business and willing to answer. The agencies that pitched their way into trouble are not in any of these samples.

Several sources sell to the firms they study. BTI sells client-feedback programs to law firms. Setup runs agency searches. Their direction agrees with the independent work; their exact figures deserve less weight.

Gardner's data comes from law firms. That it carries over to agencies and strategy firms is my inference, supported by the referral data but not measured directly.

This needs a book. A firm with no clients cannot grow inside one. For a new firm the pitch or cold outreach may be the only door available, and nothing here argues otherwise.

The pitch will always be there, and sometimes it is the right call. The work that finds you next is being decided now, in accounts you already run, by people deciding whether you were worth recommending.

Related from Sound Decisions: Nobody Pays You Extra for Being Honest · The Marketing Job Is Derived Power · The Ground Nobody Was Defending

On Didn't Sign Mbappe. They bought a category.

Most athlete deals are awareness machines. They generate impressions, not growth, and the peak athlete is the most expensive version. On's deals are structured differently, and the evidence for that distinction is worth around half a billion dollars. The question is whether the model translates from running shoes to football boots.

When On announced their entry into football alongside Kylian Mbappe this week, the instinctive read was that a Swiss running brand had bought an awareness machine at peak price. At his zenith. At Real Madrid. Nowhere to go but down in terms of visibility.

That reading is how most athlete deals work. It is not how On has operated for the last seven years, and understanding the difference is what separates a tactical critique from an actual analysis of what just happened.

I came into this research with the critique ready. Signing athletes generates impressions. It does not fundamentally drive growth. Nike loses Mbappe and gets the money back to sign Lamine Yamal and five others at better expected value. The standard argument is strong, and I pre-registered it before looking at anything.

Two of my predictions were wrong in ways that changed the piece.

Mbappe plays for Real Madrid, who have a kit deal with Adidas worth a reported €110 million per season, recently extended to 2034. My prediction was that this would prevent him from wearing On boots in competitive matches — a structural block that would reduce the deal to lifestyle content and training footage.

That was wrong about how football works. Club kit deals and personal boot sponsorship are entirely separate categories. The kit deal covers jerseys, shorts, socks — worn by the squad because the club supplies them. Boot deals are personal to the player, negotiated independently, and have nothing to do with the club's kit supplier.

Mbappe wore Nike boots at Real Madrid for two full seasons while the club trained and played in Adidas kit. Gareth Bale wore Nike boots at Adidas-kitted Real Madrid. This is standard practice across European football. There is no conflict, and On can put Mbappe on a pitch in the Champions League from day one.

That matters because the boot is the product. Not the jersey. The product story is live from the first competitive match.

I framed the deal as an endorsement and critiqued it as one. It is not an endorsement.

Federer did not endorse On. He invested in them before the partnership was announced, worked with their design team on new products before anything went public, took a reported three percent stake, and became the brand signal that helped them cross from running into tennis and then lifestyle. On went public in September 2021 at a valuation of roughly $11 billion. Federer's return on his initial position has been estimated at approximately 2,480 percent.

Pre-announcement reporting on the Mbappe deal, later confirmed accurate, described the same structure: equity shares and his own sub-brand within On's football line, not a standard endorsement fee. The press release language is worth reading literally — Mbappe will “work directly with On's product teams, integrating his elite perspective into the development and testing of future football footwear.” Thierry Henry, who is announced simultaneously as Director of Football, has been operating inside On since late 2025.

This is not a marketing department activation. This is the Federer model applied to football, built over the better part of a year before anyone outside the company knew it was happening.

On's trajectory since 2019 follows a legible pattern. Start in running, where the technology story — CloudTec®, the patented cushioning system — creates genuine product differentiation. Use running credibility to enter adjacent categories. Use the athlete model not to buy awareness but to buy category permission and co-design input from the highest available level of each sport.

On's 2026 revenue is projected at roughly $3.8 billion, growing at roughly 23 percent. Their current football revenue is zero. The global football boot market is approximately $4 billion annually, with Adidas holding around 35 percent and Nike around 30. Getting to ten percent of that market — what Nike built over decades — would add $400 million to On's top line.

That is what they are buying. Not the Mbappe impression. The football category.

Two things I came in expecting to find still hold after the research, and they are the reasons this is not a simple success story.

The second surviving critique is the one that the press release cannot answer, because no product exists yet to answer it. On's LightSpray™ technology — a robotic manufacturing process that creates a continuous sprayed filament upper — is plausible as a football innovation in theory. It is extremely light, precise in fit, and close to the foot. Whether it can handle the mechanical demands of elite football is a question being answered in a lab in Zurich as this is published. The announcement is not the result. The boot is the result, and nobody has seen it.

Nike and Adidas have decades of elite player feedback embedded in their boot construction. Their mould-fit programs are built into the culture of professional football at the club level. On is entering against two incumbents who have not been standing still.

On's most important innovation over the past seven years was not CloudTec® or LightSpray™. It was the decision to treat the athlete relationship as a product decision rather than a marketing one. That choice produced the Federer outcome. It is the framework they are applying to football.

Whether the framework survives the translation is the only question that matters. The boot does not exist yet, and the answer is in the boot.

Related from Sound Decisions: The Memory Isn't Yours · Poppi Didn't Go Viral by Accident · Grüns' Decisions Before Growth

Change the Question, Change the Ranking

Ask an assistant the same buying question two normal ways and the brands it recommends barely overlap. The number on your AI visibility dashboard is mostly a property of the tracker, not of your business.

I went into this research expecting to find that AI visibility tools count brand mentions and nothing else. That the whole category sold a tally, the tally was thin, and the thinness was the story. I wrote that down before I searched, because a prediction you record in advance is the only kind that can embarrass you.

It came back false on the first search, and it was the prediction I held most confidently. The serious tools in this category separate a citation from a recommendation, on the explicit reasoning that a brand can be quoted as a source without ever being put forward as a provider. They track sentiment. They track position. They flag when an assistant describes the brand incorrectly. The sophisticated version of this product exists, and it is not expensive.

So the problem is not that the instrument is crude. The problem is what the instrument is pointed at. And if you are paying for one of these tools, or deciding whether to, that is the part that decides whether the money is well spent.

The same question, asked twice, two different shortlists

In May 2026, a team led by Will Jack published a study of production retrieval-augmented commercial recommendation. They ran roughly 6,000 paraphrase trials against roughly 6,000 same-prompt rerun controls on OpenAI and Anthropic models, and measured how much the recommended brand sets overlapped.

Rerunning an identical prompt produced sets that agreed about half the time. Rephrasing the same buying intent in a second natural way produced sets that agreed far less. A cosmetic rewording, "best CRM" against "top CRM", landed at 0.288. Adding a constraint a real buyer would add, "best CRM for a SaaS startup", dropped it to 0.135.

The authors do not leave the implication to the reader. Counting brand mentions across a fixed prompt set, they write, produces a metric whose dominant source of variance is which paraphrase the tracker happened to issue, rather than the model's behavior toward the brand. The prompt string, not the buyer's underlying intent, is the main input to which brands surface. They name this a direct challenge to a popular practice, and they are right to.

Read that once more with your own dashboard in mind. The thing moving your score most is a choice your vendor made about wording.

What you are actually paying to watch

Every measurement system has a division of labor between what the instrument decides and what the world decides. A good one keeps the instrument's contribution small. Here it runs the other way.

The left column moves the score. The right column moves the business. The report hands you a single percentage and lets you assume it came from the right.

Two assistants, one brand, different answers

Nobody agrees how big this channel is either

Go looking for how much traffic AI assistants actually send, and you find a settled-sounding consensus that falls apart on contact.

Contentsquare's 2026 benchmark, built on 99 billion sessions across roughly 6,500 sites, puts AI-referred traffic at 0.2% of total visits, growing 632% year over year. Its own summary of that pair is the honest one: 0.2% is small and 632% is not.

Conductor's benchmark, across 1,215 enterprise domains and 3.3 billion sessions, puts the same quantity at 1.08%, with information technology reaching 2.80%.

Both are sound. They are answers to different questions being reported as one, and the gap between them is wider than most of the quarterly movements companies act on. The same collision shows up in conversion. Adobe measured AI-referred visitors to US retail sites converting 54% better than everyone else in May 2026; a year earlier the same measurement ran roughly half as well. A relative premium and an absolute rate are not the same object, and putting them on one slide produces a strategy that cannot be right.

Where this argument is weak

The load-bearing source is a preprint. It covers a specific set of models in a commercial retrieval setting, and Jaccard overlap on sets is a demanding measure that reads harsher than a looser one would. The authors report clustered bootstrap confidence intervals and say plainly that per-prompt precision is wide; the strength comes from averaging across many prompts, not from any single one.

The traffic evidence cuts against me as well as for me. A meaningful share of what analytics calls direct traffic is assistant-driven traffic that lost its referrer when somebody pasted a link into a fresh tab, and answers rendered inside a search results page are not separately attributed at all. The real footprint is larger than the low figures show. Small is not the same as ignorable, and a channel that converts well while measuring badly is the hardest kind to govern.

And the claim I most wanted to prove, I cannot. There is no public evidence that mention share fails to predict revenue. There is none that it succeeds, either. Nobody has run it, including the firms selling the dashboards. That is an open question, not a verdict, and anyone answering it confidently in either direction is selling something.

One more belongs here. Almost every non-academic source on this topic is published by a company that sells the category. Both findings this piece rests on came from arXiv, which is the only reason I trust them more than the rest.

The Buyer Wrote Down What They Bought From You. And they said why.

Every entry in this series so far has been written from the founder's chair. This one is written from the buyer's, because the assumptions in their model are the ones your price was built on — and years later they publish which ones were wrong.

If you are preparing to sell, you have probably spent months on the story: the growth, the margin structure, the loyalty of your buyers, the runway of the category. You could tell it in your sleep. What almost nobody does is the reverse exercise: work out what the acquirer's model has to assume for the number on the table to make sense.

Those assumptions are the deal. The price is just their arithmetic made visible.

And there is an unusual feature of this particular market: the buyer eventually publishes which assumptions were wrong, in a document anyone can read, with the reasons broken out line by line. Almost nobody reads it.

When a company acquires a brand, the price splits. Part of it attaches to things you can point at — plants, inventory, receivables. The rest becomes goodwill and indefinite-lived intangible assets: the accounting name for the portion of the price not attributable to any identifiable thing.

That portion is not amortized. It sits on the balance sheet at full value until the company tests it and concludes it is worth less than it is carried at. Then it is written down, and the write-down is a public statement about what the buyer thinks they actually got.

It was not an isolated line. The same interim test found seven of twenty goodwill reporting units and six brands carried above their fair value. Total non-cash impairment was $15.4 billion — $7.1 billion of goodwill and $8.3 billion of indefinite-lived intangibles. Aggregate goodwill and indefinite-lived intangibles fell from $97.3 billion to $80.5 billion in a single quarter. Oscar Mayer was impaired again in 2020, another $626 million.

Here is the part worth the price of admission. Kraft Heinz did not simply record the charge. Its 10-K sets out the drivers brand by brand, and for Oscar Mayer they were: sustained increases in supply chain costs, expectations for lower pricing to maintain competitive positioning, and expectations for increased marketing investment.

Read that as a founder rather than as an analyst and it is a confession about the original model.

The middle row is the one that matters to anyone who is about to sell. The acquirer's model assumes your brand holds its price with less support than you are currently giving it. That is not a cynical reading; it is the arithmetic of most strategic acquisitions. The synergy case is largely a support-reduction case. If the brand needed everything you were spending, there would be no synergy to pay you for.

So the question a founder should be asking in diligence is not what the buyer will pay. It is: how much of what I currently spend do they believe they can stop spending, and are they right?

There is a structural reason the answer surfaces late, and it is not incompetence.

Goodwill is not amortized. It sits at full carrying value until an impairment test says otherwise, which means the cost of overpaying does not touch earnings on any fixed schedule. Research published in the Journal of Accounting Research in 2025, estimated across 860 all-cash takeover auctions between 2001 and 2022, found that this treatment measurably raises what strategic bidders are willing to pay — because the acquisition cost relating to goodwill will not affect earnings until an uncertain later date when an impairment occurs.

That finding cuts two ways and a seller should hold both. The deferral is part of why the offer on your table is as high as it is. It is also why nobody, including the buyer, has a strong incentive to test the assumption carefully before signing.

The same research puts a figure on the distortion: remove the impairment-only treatment and the share of private equity acquirers rises by roughly 6.9 percentage points. The accounting convention changes who buys companies.

This series has argued before that a plan tells you what a company hopes and a rule tells you what it will do when hope runs out. An acquirer's stated intention for your brand is a plan. If it matters to you, it belongs in the document.

Four limits, and the first is the one I had to check before writing anything.

Impairments are heavily driven by discount rates. The Federal Reserve's tightening cycle from 2022 lifted the cost of capital across nearly every industry and pushed borderline reporting units into impairment territory for reasons having nothing to do with deal quality. Anyone reading a write-down as proof of a bad acquisition is overreading. What survives that objection is narrower: Kroll's 2026 study found 2025 impairments concentrated in healthcare, consumer staples and industrials — in a year when rates were falling and aggregate impairment volumes were moderating. A rate story would have moderated with everything else.

This is one company, examined closely. Kraft Heinz is the most documented case available because the charge was enormous and the litigation made the record public. It is not a sample.

I expected the disclosure to be worse than it is. Before looking, I predicted no acquirer would publicly reconcile a write-down to its original thesis. That was wrong. The filing does it brand by brand, in plain language, with the drivers named. The information problem here is not availability. It is that nobody reads 10-K brand impairment disclosures while preparing to sell a company.

A write-down is not a verdict on the seller. Oscar Mayer's founder did not do anything wrong in 1883. The charge says the acquirer's model was optimistic, not that the asset was bad.

Sellers spend the diligence period being examined. Very few of them examine back, and the material to do it with is sitting on a public filing server, indexed, searchable and free.

The buyer will eventually publish what they thought they were purchasing and what they turned out to have. The only choice available is whether you read that document before the deal or after it.

Related from Sound Decisions: Purely Elizabeth Took Eight Years to Raise $3 Million · Can Your Mission Survive Acquisition? · The Rule That Never Moved

ESPN Needs 30 Million People to Pay for It. Nobody knows how many do.

For three decades the bundle meant ESPN never had to find out what it was worth to anyone in particular. That question is now the only one that matters, and the company stopped publishing the number that would answer it in the same month it started asking.

If some of your revenue arrives because you are part of something bigger (a bundle, a default setting, a contract nobody reopens), you know it doesn't feel like a moat. It feels like the business. The money comes in every month, the renewals go through, and nobody has to decide to choose you, which is exactly why it is so easy to read as loyalty.

ESPN ran on that kind of revenue for thirty years. What happened when it stopped is usually told as a story about technology.

The standard account of ESPN's decline goes like this: cord-cutting eroded the subscriber base, rights costs exploded, social platforms commoditized highlights, and the company pivoted to streaming to meet the audience where it now lives. Every clause is true. Together they describe the weather rather than the decision.

The decision underneath is older and simpler, and almost every version of this story gets it backwards.

ESPN is usually described as a company whose competitive advantage was its programming, now lost to fragmentation. That reads the business the wrong way round.

ESPN's advantage was that it was mandatory. It sat in the basic tier of essentially every cable package in America, and it collected a fee from every household in that package whether or not a single person in the house had ever watched a game. Analysts at SNL Kagan put the carriage fee at $8 to $9 per subscriber per month, the highest of any basic cable network by a wide margin.

That is not a product advantage. It is a distribution advantage, and it was the most durable one in modern American media, because it converted indifference into revenue. The people who never watched ESPN were, in aggregate, one of its largest customer segments.

When Disney broke out sports segment results for the first time, the shape of it became public: $10.8 billion in affiliate fees, against $4.4 billion in advertising and $1.1 billion in ESPN+ subscriptions. Affiliate fees were roughly two-thirds of the business, and they arrived without anyone choosing to buy anything.

Thirty million households that actively choose to pay, every month, in place of a hundred million that were never asked.

The useful frame here is not product versus distribution. It is a single question: is the person who pays the same person who chooses?

Every symptom in the standard account descends from that row. Rights inflation hurts because the base paying for them is shrinking. Highlight commoditization hurts because highlights were the retention wrapper on a product nobody was choosing. Debate programming exists because it is cheap to fill the hours between the games that actually justify the fee.

Here is the part I did not expect to find, and it changes the story from the one I set out to write.

ESPN's subscriber base has fallen from roughly 100 million at its 2011 peak to an estimated 55 to 58 million in 2026. Yet Disney's own fourth-quarter fiscal 2025 disclosure describes subscription and affiliate revenue increasing, reflecting higher effective rates, partially offset by fewer subscribers. The same sentence appears for the international business.

Revenue is being held up by charging more per remaining subscriber.

Anyone who read the piece on reformulation in packaged food will recognize the move exactly. National brands grew dollars while losing units, covering volume decline with price, on a shrinking base of buyers. It is the same manoeuvre, in a different industry, with the same property: it works until it doesn't, and every year it works makes the following year harder, because the price rise is itself a reason to leave.

So the demand test has not arrived. It has been deferred, and the instrument of deferral is a price increase on the households that have not yet left.

This is not unique to sports. The same wall went up across packaged food inside a single year, as three of the five largest breakfast operators were acquired or absorbed and stopped reporting separable US numbers. When a business becomes difficult to explain, the reporting granularity tends to fall first.

There is a second reason the transition is harder than it looks, and it is architectural rather than financial.

A streaming service is four layers running simultaneously. Data: who is watching, on what, at what bitrate, from where. Business rules: entitlements, tiers, concurrency limits, and the geographic blackout logic that live sports rights demand. Optimization: personalization, ad yield, churn prediction. Interface: the thing a person actually touches at kickoff.

In the bundle, ESPN operated none of them. The cable operator owned the interface, the entitlements, the billing relationship and the customer data. ESPN made programming and collected a toll. It has now been handed all four layers at once, with no operating history in any of them, while the live-sports failure mode is unforgiving: nobody tolerates a buffering wheel at fourth and goal, and a blackout error screen delivered to a paying subscriber is a cancellation with a delay on it.

We have one clean test of how that goes, because ESPN recently tried to take a fifth layer by force.

Three things, and the first is substantial.

Disney says the transition is working. On the second-quarter fiscal 2026 call, the company stated that the new direct-to-consumer service was generating more revenue from digital subscribers than it was losing to declines in traditional television customers. If that holds, the replacement is running ahead of the erosion and the arithmetic above is pessimistic. It is also, by construction, unverifiable from outside — the disclosure that would confirm or refute it is the one that was discontinued.

The thirty million figure is mine, not ESPN's. It divides a disclosed affiliate number by a list price. It ignores advertising revenue, which is substantial and travels with the audience rather than the bundle; it ignores bundled pricing, promotional tiers and the NFL Network assets ESPN has since acquired; and it treats a 2023 disclosure as a stable baseline when affiliate revenue has moved since. Read it as the order of magnitude, not the target.

Sports may simply be different. Live rights are the last appointment viewing in the culture, and the willingness to pay for them may be far higher than for general entertainment. If any category can convert a hundred million indifferent households into thirty million committed ones, it is this one.

The bundle was not merely a distribution channel. It was an arrangement that made a question unnecessary for thirty years, and ESPN built an extraordinary business inside the space where that question would otherwise have been.

The question has now been asked. What is remarkable is not that the answer is uncertain. It is that the company chose this moment to stop publishing it.

Related from Sound Decisions: If You Take the Quality Out · Walmart Can End the Company · The Rule That Never Moved

A Third of the Shelf. Two percent of the growth.

Roughly a dozen major consumer companies are running the same restructuring playbook simultaneously, and most are on their second or third round. One restructuring is a correction. Four in a row is a system that cannot locate its own constraint.

You know the pricing meeting at a big brand. The increase is on the slide before anyone has argued for it, because it has gone through every year. The share numbers are steady, velocity is holding, price realization is on plan. Nobody in the room is asking whether the brand gave anyone a new reason to choose it this year, because nothing on the slide is built to ask.

Then look up from the slide at what the biggest companies in the aisle have announced. Nestlé announced 16,000 job cuts. Kraft Heinz reversed a $28 billion split and restructured instead. PepsiCo closed five plants and is removing about a fifth of its US products. General Mills shut three plants and is running a restructuring program to fiscal 2028. Unilever is exiting ice cream. Procter & Gamble is absorbing $1.6 billion in restructuring costs. Whirlpool has restructured more or less continuously since March 2024.

The proximate causes are well covered and all real: inflation, tariffs, GLP-1 adoption, a weak housing market, private label. What the coverage does not explain is why the same playbook is appearing across unrelated categories at the same time, or why the same companies keep running it again.

Those two facts — simultaneity and repetition — are the ones worth a structural explanation, because neither is what a category shock looks like.

Read the three figures together and the shape of the problem is visible without any interpretation. Private label and small independents are taking 83 percent of the growth between them. The companies that own the shelf are taking almost none of it.

And small brands were not always there. They held 13 percent of the US market in 2021 and delivered 15 percent of growth. By 2025 they were delivering 35 percent. The share barely moved. The growth capture more than doubled.

Private label wins on price, small brands win on function, and large brands are caught in the middle.

McKinsey & Company, State of Food & Beverage, April 2026

That sentence is the whole diagnosis, and the consumer data underneath it is specific: 37 percent of US and UK consumers name functional benefits as their main reason for buying a small brand, against 18 percent who say the same about a large one. Eighty-six percent rate private label as equal to or better than branded on value.

The architecture of a major consumer company rests on an assumption: that scale confers pricing power. National distribution, mass reach, shelf dominance, and the annual price increase that follows from all three.

Three things broke that independently, and none of them is about consumer taste.

Distribution scale became rentable. Brokers, Amazon, and retailers actively hunting differentiation will put a four-person brand on a national shelf. Media scale became divisible. Attention is sold in micro-units, and a small brand buys the same eyeballs per dollar or better. Manufacturing scale became a service. Co-packers will make your product at a quality and cost a mid-size company could not have reached alone in 1995.

So a portion of what these companies charged was never a cost function. It was a position — the ability to be the thing that was there when nobody was choosing. When that position erodes, cost is the only lever left, and cost is a fight with private label on private label's terms: identical cost base, no marketing overhead, and the retailer's own shelf to stand on.

This is the part that turns a trade summary into a decision.

The instruments read healthy. These companies measure share, velocity, distribution and price realization to a professional standard. Every one of those measures a market structure. None of them measures why someone chooses you. So when the underlying reason erodes, the dashboard stays green until the moment price has to move — at which point the problem presents as a pricing problem, and gets a pricing answer.

The boundary nobody wrote. Everyone took price through the inflation window because the window allowed it. A pricing executive who works with several of the majors described the subsequent retreat publicly as a strategic reset rather than a reversal, noting that shopper elasticity and retailer pushback only became apparent after several years of inflation-driven pricing. There was no rule that would have failed that decision, because the decision was correct at every layer where a check existed. It made the quarter.

The rule that should have stayed a judgment. The annual price increase stopped being a decision and became a default. A default does not ask whether the brand added anything that year.

Closing plants, cutting products and lowering price are surface corrections to a fault two layers down. They restore margin without restoring the reason to be chosen. So the reading drifts again, and the next round gets announced.

Registered before writing, and two of them are live problems rather than hypotheticals.

It is not universal, and the exceptions are instructive. Of 166 food categories where store brands compete, 52 percent saw private label unit growth. Forty-eight percent did not. In the breakfast occasion specifically, private label is declining faster than the majors — down 4.7 percent in ready-to-eat cereal against a segment down 1.0. Anyone applying the private label story as a universal law is overreaching, and that includes anyone reading this.

The price concession is not yet in the aggregate data. National brand dollar sales rose 2.2 percent in the first half of 2026 while units fell 0.5 percent. The announced cuts at PepsiCo and General Mills came in December 2025 and February 2026 and are too recent to show. Until they do, the majors are still taking price, not conceding it. That is a real gap between the announcements and the evidence.

The four-quarter test cannot be run yet. If the companies cutting price recover share durably within four quarters, then pricing power was intact and merely mispriced, and this argument is wrong. That will be checkable in 2027.

The measures do not agree with each other. Circana has reported private label at 25.5 percent of total unit sales in 2023 and at a record 23.8 percent of grocery unit share in 2026. Both are Circana figures on different universes. Anyone quoting one without naming which is producing a confident wrong comparison.

One thing this is not: an argument that the operators are incompetent. It is stronger if they are excellent, because that makes the failure architectural rather than personal.

The majors conceding price is not a story about the majors. It is a forecast about every premium brand standing near them, and the timeline is short.

The question stops being strategic and becomes operational. Which half of your premium is specification, and which half is identity? Nobody in the category has an instrument that separates the two, which is why the answer keeps arriving as a surprise, one restructuring at a time.

Related from Sound Decisions: ESPN Needs 30 Million People to Pay for It · If You Take the Quality Out · Distribution Doesn't Create Demand

The Rule That Never Moved. It isn't what anyone gets paid on.

The plan is what a company intends. The rule is what it actually is. Only one of them shows up in a bad quarter, and it is usually the one nobody is compensated for.

Somewhere in Minneapolis last spring, a finance team closed a fiscal year that had gone almost entirely against them. Sales down. Profit down. A goodwill and brand intangible write-down large enough to flip the operating line. Every headline number a company would rather not print, printed.

And in the same annual report, one sentence saying the year's productivity objective had been met in full.

That is the sentence worth reading. Not because it rescues the year, because it doesn't, but because of what it survived.

In plain language, HMM is a standing obligation. Every year, the business must generate enough productivity savings, mix management and price realization to absorb its own input-cost inflation, hold its margins, and still throw off money to put back into the brands.

The delivery has stayed as well. Roughly 6 percent of cost of goods sold in fiscal 2024. Around 5 percent in fiscal 2025. Five percent again in fiscal 2026 — the year of the write-down.

Everything else moved. Fiscal 2025's stated top priority was accelerating organic sales growth, and specifically volume. By the fiscal 2027 outlook the language had become a step change in remarkability and an acceleration of enterprise transformation, alongside a target of three billion dollars in cumulative cost savings by fiscal 2030. Different words, different emphasis, a different diagnosis of what was wrong.

The narrative changed every year. The rule didn't change once.

Here is the part that turns an observation into a finding. HMM is not what the executives get paid on.

That is the shape of a real constraint, and it is worth sitting with. The bonus metrics are outcomes. They depend on consumers, competitors, retailers and a dozen things nobody in Minneapolis controls. HMM is a boundary. It does not ask the business to predict the year correctly. It asks the business never to let inflation eat the margin without a fight, whatever the year turns out to be.

One requires being right. The other requires only holding the line.

Most planning is positive. It says what the business will do: enter this channel, launch that line, grow at this rate. Positive plans are necessary and they are also fragile, because every one of them is a prediction wearing a deadline. When the world changes, the plan is wrong, and the organization spends the next quarter arguing about whose fault that is.

A negative rule makes no prediction. It states what the business will never allow — never take the promotion below this yield, never carry a customer below this margin, never let a category fund its growth out of the brand's price integrity. It cannot be wrong about the future because it is not making a claim about the future. It can only be enforced or abandoned.

And that is the second reason to look for the invariant rather than the plan. A plan tells you what a company hopes. A rule tells you what it will do when hope runs out.

Every company has one. Most have never written it down, which means most companies discover their real constraint the same way — retroactively, in a bad quarter, by finding out what they were unwilling to give up when something had to go.

This matters more than the tidy version of the story. Anyone selling you a framework that promises otherwise is selling you a forecast with better manners.

The company itself appears to read it this way. The response to the bad year was not to relax the rule but to add to it — a larger, longer cost-savings target running to fiscal 2030 — while attacking demand with a separate and differently-named effort. Boundary on one layer. Judgment on another. That separation is the actual lesson, and it is available to a business a thousandth of the size.

Find out which is which before the year decides for you.

The plan is what you intend. The rule is what you are. Only one of them shows up in a bad quarter.

Related from Sound Decisions: Hope Is Not a Strategy · Most of Your Marketing Budget Is Rent · Nobody Pays You Extra for Being Honest

The Fear Tariff. The budget cut arrives before the cost does.

The last time a tariff shock hit, the marketing cut came months before any cost did, and the recovery overshot. Canada's counter-tariffs are now in force and the U.S. has escalated. The cut that matters this quarter is probably not the one anyone is watching for.

The call usually comes before the numbers do. A client in a category that has been in the tariff headlines says they need to pull back spend for the next quarter, maybe two. You ask what the tariffs are costing them. There is a pause, and then some version of: we don't know yet, but finance wants us to be careful.

If you are the one taking that call, it is hard to push back. The client is nervous, the news is loud, and arguing for spend in a trade war sounds like arguing for your own retainer. This piece does not forecast where the current round goes. It describes what marketing budgets did the last time a comparable shock hit, because that pattern is the one agencies are now inside.

The obvious story is that a tariff raises costs, margins compress, and marketing gets cut in proportion. The way that story is wrong is the finding.

After the U.S. tariff announcements of early 2025, forecasters cut before any added cost had reached an income statement. By June 2025, WPP Media had revised its 2025 global ad growth forecast down from 7.7 percent to 6.0, citing disruption to global trade. By December, with tariffs still in place, it raised the forecast to 8.8 percent, above where the year had started. Its head of business intelligence, Kate Scott-Dawkins, said plainly that tariffs had not had the impact expected; retailers pulled inventory forward and consumer spending held up.

A finance chief facing a trade shock cannot cut headcount by Friday. Rent and most contracts are locked. Marketing is different: media can be paused with a phone call and a retainer frozen before renewal. It is the one large expense line that reacts at the speed of a headline, which is why the cut shows up before the cost does.

This is not unique to tariffs. The CMO Survey, fielded in January 2026 among 308 U.S. marketing leaders, found that when profits fall short, 53.1 percent of executives focus on cutting expenses, and marketing is the expense cut 45.4 percent of the time, more than any other category. Tariffs do not invent that pattern. They trigger it.

The cost-driven effect is real, but it arrives later and is narrower. In the same survey, after a year of live tariffs, 60.6 percent of companies reported no impact on business investment. Of those that changed course, cuts outnumbered increases almost four to one. That question asks about investment broadly, not marketing specifically, so it is a proxy, but the shape is consistent: a fast, broad reaction on fear, and a slower, narrower one on cost.

Exposure follows category and supply chain, not company size. Two clients one desk apart can have opposite years: one in a directly named category with nowhere else to source, one outside every list, cutting on a headline borrowed from someone else.

What has happened since this piece first ran on September 7, the day before Canada's counter-tariffs took effect. Canada's counter-tariffs took effect at 12:01 a.m. on September 8, covering $27.6 billion of U.S. imports at rates of 15, 25 and 50 percent, matching the U.S. Section 338 tariffs of August 22. The same day, the U.S. signed further Section 338 proclamations that move certain Canadian motor vehicles, dairy and alcoholic beverages from tariffs to outright import bans from September 29.

This is a pattern seen once, not a forecast. The 2025 round resolved more gently than feared; nothing guarantees 2026 does. Section 338 actions carry no fixed end date, and this round has already escalated to import bans.

The 2025 rebound had two causes. WPP Media credited milder tariff outcomes and the AI investment boom. Relief alone may not repeat the overshoot.

The CMO Survey investment question is broad, covering business investment in general rather than marketing, and it samples U.S. companies. It is the clearest available proxy, used as one.

Forecasters are not advertisers. A forecast revision shows expectations moving, which is what drives budget decisions in the moment. It is not the same as measured spend.

The tariff lists were public weeks before they took effect. None of this required a forecast. It required someone to check each client's category before the client called asking what to do.

Related from Sound Decisions: When the Money Changes Its Mind · Your Best Clients Are the Ones You Undercharged · You Don’t Need More Clients. You Need Longer Ones.

Starbucks Ran the Numbers on Its Own Turnaround — and the prediction I sealed in advance came back wrong

I built an instrument to classify a company's growth moves as widening the gap between what customers pay and what employees and suppliers accept, or just shifting it. Starbucks was the first real company I ran it against. I predicted one outcome, sealed it in writing, and got it wrong by a factor of two.

A few years ago, right after COVID, I sat down in a Starbucks in New York and couldn't hear myself think. The music was loud enough that sitting still felt like the wrong thing to do. Nobody was going to linger over a laptop or a conversation in that room. The place that once sold itself as a "third place" — somewhere between home and work, worth showing up to even when you weren't buying much — had quietly become a room optimized to get you out of it.

That's a company making a decision, over and over, to take more out of a relationship than it puts in. And it almost never feels like a bad decision from the inside. It feels like good management: turn tables faster, protect margin, hit the quarter. Short-term incentives rarely announce themselves as short-term. They just show up as the obviously correct call.

A bigger pie, or a bigger slice

That's the actual question behind Sound Decisions — not "good decisions versus bad" in some vague moral sense, but a narrower and more useful one: is a business, in a given decision, making the pie bigger, or just taking a bigger slice of the same pie? I'd call the first value creation — raising what the product is worth to the person paying, or lowering what it costs the people supplying it, or both. I'd call the second value extraction — moving money from one side of the table to the other without changing what's on the table.

This isn't a moral argument, on purpose. It's an arithmetic one. A business that widens the gap between what customers will pay and what employees and suppliers will accept has more total value to work with, compounding, every year after. A business that only manages the split of a fixed gap is optimizing a number that can't grow on its own. This is the founding thesis of CULT+MATH, the firm behind this publication. So instead of opening with the framework, I wanted to open with a real company and see if the idea survives contact with a real balance sheet.

The instrument, and its limits

I built an instrument to make this classification rigorous instead of vibes-based — the Net Value Position. Feed it a company's actual growth decisions over a defined window and it sorts each one: did this move widen the gap (Creates), just shift the split (Divides), do a bit of both, or do neither.

Starbucks was the first real company I ran it against, for a specific reason: it's one of the few companies where both halves of that gap are visible from the outside. Most companies make the employee side invisible. Starbucks, thanks to a very public union campaign, strike coverage, and turnover disclosure, doesn't get that luxury.

To be precise about scope, because a decision aid that overclaims is useless: this run covers March 2025 through September 2026, under CEO Brian Niccol. It does not classify the louder, emptier era I opened with — that's a personal observation, not something the instrument tested. What follows is sourced only from filings, earnings calls, and dated company statements.

The prior was the problem, not the taxonomy

Before looking at a single source, I sealed a prediction: net dividing, with price increases leading the dividing column, because coffee costs had risen almost a dollar a pound plus tariffs, and I believed — stated as a bias, in advance — that large restaurant operators default to splitting the gap because price is the fastest lever a public company has under quarterly pressure.

There were no in-window list-price increases. Not one. Faced with the fastest lever available, Starbucks declined to pull it.

The honest position: a creating strategy on the customer side, financed in part by a division on the employee side the company hasn't settled. Not a clean story. A specific one, which is more useful. I could have quietly adjusted my prior after the fact and written this as if I'd expected it all along — that's the easier essay, and it's exactly the kind of self-serving edit that makes an instrument untrustworthy. The prediction was sealed before I looked. It was wrong, in a way that reveals something: I assumed a company under margin pressure reaches for price. This one didn't.

Build It Anyway

Nobody is coming to look at it.

A note. No research behind this one — no number, no case, no sourced claim. The rest of the series is built on evidence; this is not, and it says so here so you can decide whether to keep reading.

Not today, and probably not the day after you finish. You will ship the thing you have been carrying for eleven months and the number of people who notice will be smaller than the number of people you told about it. Some of them will say congratulations. Two will mean it. Then everyone goes back to their own eleven-month thing, because they have one too.

This is not cruelty. It is arithmetic. Attention is finite and yours is already spent on your own work, which is exactly where theirs is.

And if it fails — which it probably will, because most things do, and the base rate does not make exceptions for how much you wanted it — there will be no funeral. No one will write about the version of you that put four years into something the market shrugged at. The people who love you will be kind about it, in the way people are kind about a haircut. Then Tuesday will happen. Then another Tuesday.

That is the whole situation. It is worth looking at directly, because the two most common responses to it are both lies, and both of them let you stop.

The first is that it will be worth it.

That one says: keep going, because there is a payoff waiting that will retroactively justify the cost. The exit. The recognition. The moment someone finally sees it.

Sometimes that arrives. Mostly it does not, and when it does it is smaller than advertised and lasts about a week. If your reason for building is a future ledger that balances, you have staked the work on an event outside your control, and the day you stop believing in the event is the day you stop.

The second is that nothing matters, so why bother.

This one sounds tougher and is actually the softer of the two. It takes a true premise — the universe is not keeping score — and derives a false conclusion, which is that you may as well not. It is the lie that lets you quit while feeling clear-eyed about it. Apathy wearing realism as a costume.

Both lies do the same job. They give you a reason to put it down.

It is not inspiring, which is how you know it might hold.

The universe is not keeping score. Correct. You are.

The thing in front of you today is either done well or it is not, and that difference is real regardless of whether anyone ever audits it. The paragraph is clear or it is muddy. The number is checked or it is guessed. The customer is better off or they are not. None of that requires cosmic ratification to be true. It is true at the resolution where you are standing, which is the only resolution you were ever going to work at.

So the question stops being will this matter and becomes is this good. That is a smaller question. It is also one you can answer today, alone, without permission, and it does not depend on anybody showing up.

The old story has a man pushing a rock up a hill forever, and the punchline is that it rolls back down every time. Everyone remembers the futility. Almost nobody remembers the last line, which is that we have to imagine him happy.

Not resigned. Not brave. Happy — because the rock is genuinely heavy and he is genuinely strong and the hill is genuinely there, and none of that stops being real just because the hill does not care.

Build it anyway. Not because it will be remembered. Not because it will work. Because the alternative is to spend the same finite Tuesday not building it, and you will still be you at the end of that Tuesday, and the rock will still be at the bottom of the hill.

Might as well carve it properly on the way up.

Related from Sound Decisions: Nobody Pays You Extra for Being Honest · Hope Is Not a Strategy · Leaving With Something

The Budget Line. Optimism is up. Investment isn't.

More than two-thirds of agencies expected this year to beat last year. Fewer than half expected to invest in it. What crosses that gap is not what saves time; it is what gets a firm found, and what proves it is worth the fee.

Ask a room of agency owners whether this year will beat last year and most hands go up. Ask what they are spending to make it happen and the room goes quiet.

That gap is measurable. RSW/US surveys marketing, advertising, digital, PR and consulting firms across the United States and Canada, drawn from a sample of more than 5,000, alongside a panel of senior marketers. Going into 2026, more firms expected their business to improve than two years earlier. Far fewer expected to put money behind it.

The clients were cooler still. Only 56 percent of client-side firms expected improvement, down from 78 and 79 percent in the two prior years, and 37 percent expected to invest. Just 25 percent expected their marketing spending to rise, against 57 percent a year earlier. Agencies are more confident than the people who pay them.

Almost every tool sold into this market promises the same thing: it will save hours. That may be true and still be useless against a bank account, because hours are not a line item. They turn into money by one of two routes, work you would otherwise have turned away or cost you would otherwise have carried, and a seller who cannot name which one is asking you to do the arithmetic yourself.

Agencies have lost the same argument from the other side for years. Strategy comes off the estimate not because clients think it worthless, but because it arrives with no budget attached. The test runs both ways in one sentence: name the budget this comes out of, and what stops being bought.

If most firms are not investing, what does the other half fund? The clearest answer in the data is being found, because the routes firms were found by are shrinking.

Behind visibility sits margin, bought indirectly. Nobody purchases margin; they purchase what stops it leaking, and the largest leak is value nobody has written down. The 4A's and the ANA reported in June 2024 that only 10 percent of agencies and 5 percent of clients have a formal definition of the value an agency delivers. Fees get negotiated over something neither side has defined, which is how a rate card becomes the only language in the room.

The same RSW/US report shows why that matters. Asked why they pass on a firm, marketers ranked weak strategic thinking second, behind lack of category experience and ahead of creative quality. Agencies report 60 percent of clients now have some in-house capability, up from 40 percent, and only 7 percent of clients say they prefer to buy productized offerings; 65 percent prefer services and capabilities.

A senior association executive who spends his weeks with independent agency owners gave the same ranking when asked, unprompted, what they will actually spend on this year: visibility first, profitability second, and everything else fighting for room. That is one experienced observer's impression, not a measurement. It points the same way the survey does.

These are intentions, stated in late 2025, for a year now three-quarters gone. Read them as the mood firms planned under, not as a record of what they spent.

RSW/US sells outbound new-business services and runs an agency-search business, and its report recommends investing in exactly that. The sample is drawn from more than 5,000 firms; the number that responded is not published.

Visibility is evidenced by the channel data. Profitability as a purchase driver is not; it rests on reasoning and on one observer's read. If you trust only the measured half, trust visibility.

The web-search shift is about discovery, not selection. A client who finds you through search may still hire on referral. The data shows how firms enter the frame, not who wins.

More than half the market has decided to wait, findable only by routes that are shrinking. That is the opening. It does not require outspending anyone. It requires one funded line, aimed at how clients now find firms, while most of the market funds none.

Related from Sound Decisions: Your Next Client Is Already Inside This One · Anyone Can Look Like You Now · They’ll Pay for the Work. They Won’t Pay for the Thinking.

Nobody Pays You Extra for Being Honest. But half of them leave when they catch you.

I have been on both sides of this. First as the kid the advertising was aimed at, then as one of the people in the room deciding where it pointed. This is what the evidence says about whether doing right pays — and it is not what either side wants to hear.

I came to America at nine. My mother worked constantly — weekdays, weekends, whatever was available — and did it alone, for my brother and me. So I came home from school to an empty apartment and made myself something to eat out of whatever was in the kitchen, which is a thing a lot of children do and which nobody in a marketing meeting has ever pictured while approving a campaign.

And I watched television. Saturday mornings, and afternoons between homework and soccer.

What I want to be precise about is that I did not experience the advertising as advertising. I experienced it as information. I was a boy from somewhere else trying to work out how this place operated, and the ads were the clearest signal available — here is what a family looks like, here is what a good breakfast is, here is what a boy your age wants and eats and wears and becomes. I did not have the equipment to separate the description from the sale, because I did not know there was a sale.

So I ate the things. I wanted the things. I used them as a foothold into a culture I was climbing into, and it worked, in the sense that footholds work.

Years later I went to work in advertising, then on the client side, and eventually sat on brands I had grown up inside. I want to describe that accurately rather than dramatically: it is not a scandal. Nobody in those rooms was cackling. They were competent people with targets, under real pressure, making defensible calls one at a time. That is what makes it worth writing about. The mechanism needs no villains and would be far less durable if it did.

What changed was that I could see the machinery, and I could see who it was pointed at. I had been the target. Now I was holding the instrument.

This is the question I actually wanted answered, because the thing I believe — that a business can be honest and make a great deal of money — is worth nothing if I only believe it because I want to.

So I looked for the evidence against it first. There is more of it than I expected.

Three percent. That funnel is the most important thing in this piece, and it is the reason most writing on this subject is useless. The gap between what people say about their values and what they do at a shelf is not a rounding error. It is the whole distance. Academics call it the ethical intention-behavior gap and have been documenting it for two decades across food, clothing, travel and electronics. Price, habit, convenience and availability win, repeatedly, against stated conviction.

If your business case for integrity is that customers will reward you for it, the business case is three percent thick.

There is a famous counter to all this, and it deserves examination because it has been repeated in more decks than almost any statistic in modern marketing.

The academic literature is more careful and less quotable. Pool every study of corporate social responsibility against financial performance and the correlation is statistically insignificant, with strong evidence of publication bias — the tests that detect it come back positive. That is the honest headline and it is not flattering.

But there is a second layer. When researchers restrict the pool to studies that either avoid the confounding effects of economic downturns or use methods designed to handle endogeneity — the problem that profitable firms can afford to behave well, so the causation may run backwards — roughly 86 percent of those better-identified studies do find positive, significant effects. And the effects concentrate in consumer-facing industries.

So something is there. It is modest, it is difficult to measure, and it is nearly invisible in the raw literature. What it is not is a premium that customers pay you for being good.

Put the three percent next to a different number and the mechanism appears.

Integrity is not a product feature you charge for. It is the absence of a detected betrayal, and it is priced like insurance: you pay for it continuously, you see no return in any quarter where nothing goes wrong, and the one time it matters it is the difference between a bad year and a broken business.

That is far less inspiring than the claim made on conference stages. It is also the only version that survives a conversation with a CFO, because it does not require anyone to believe consumers are better than they are.

And it reframes the decision. The question in the room is never "should we be ethical." It is "can we take this cost out without anyone noticing." Those are the same question, and only one of them has a measurable exit rate attached to the wrong answer.

Two experiments by Luguri and Strahilevitz exposed representative samples of American consumers to manipulative interface design. Mild versions made people more than twice as likely to sign up for a dubious service. Aggressive versions made them almost four times as likely.

Three details from that work matter more than the headline. Aggressive manipulation triggered backlash and distrust — people noticed, and resented it. Subtle manipulation frequently went unnoticed entirely. And susceptibility to the subtle version was higher among less educated users.

The subtle version works better because it is not noticed. The aggressive version gets caught.

The finding, stated plainly

Now consider what a system optimized on conversion does with that. It does not choose between subtle and aggressive on principle. It tests, measures, and converges on whatever performs — and what performs is the version that slips past. Researchers describe personalization that adapts urgency cues to the individual, engages people at emotionally vulnerable moments, and exposes lower-income users to more hidden costs and dynamic pricing. The manipulation is not new. What is new is that finding the exact pressure point for a specific person is now cheap, fast and continuous.

Which means the sharp end points where it always pointed, only harder. The people least equipped to detect the subtle version are the people with the least education, the least money, the least time and the least experience of the culture doing the selling. That is not a hypothetical population to me. That was a nine-year-old in front of a television, treating advertisements as an instruction manual for how to be American.

The hopeful version of this says the same tools arm the other side. Read the label. Compare the formulation. Check the claim against the filing. I would like that to be true, and I went looking for evidence that it is, and did not find any. What has been measured is the attack.

I do still believe a business can be honest and make an enormous amount of money. I believe it more now than before I checked, but for a different reason than I used to. Not because the market rewards virtue — it barely does, three percent worth. Because the alternative carries a bill that arrives later, larger, and at the worst possible moment, and because the people who pay it first are usually the ones who could least afford the product in the first place.

I was one of them. I did not know it at the time. That is the entire point of the subtle version.

Related from Sound Decisions: If You Take the Quality Out · The Memory Isn't Yours · Value Creation Beats Extraction

If You Take the Quality Out, don’t be surprised when Walmart puts it back in.

Ask people how a company should handle rising costs and they rank the options clearly. Raising the price is the least unfair thing you can do. Quietly cutting the quality is the most. For four years the industry has been choosing the second one.

The cost meeting always arrives at the same table. Input costs are up, the price point is sacred, and someone has to choose. Three levers sit in front of you, and every one of them has been pulled somewhere in the last four years.

Raise the price. Shrink the pack and hold the price. Or change the formulation — swap the cocoa butter, cut the oil, move the cheaper ingredient up the list — and hold both the price and the pack.

The third one is the sophisticated answer. It protects the price point, protects the shelf presence, protects the trade relationship, and nobody has to explain anything to anybody. That's the case for it, and it's a real case. It is also, on the evidence, the most expensive decision available.

Ioannis Evangelidis at Esade Business School ran ten studies, published in the Journal of Consumer Research, in which consumers were told how companies had responded to rising costs and asked to judge them. The ordering came out consistently, and it is the exact inverse of how the options get ranked in a planning meeting.

Read the two together and the strategy inverts. The industry spent four years avoiding the option consumers mind least, by using the option consumers punish hardest, on the assumption that nobody would notice. Roughly half of them noticed.

Smart Balance is the visible instance of something that ran quietly across the category. Cocoa butter substituted in confectionery. Sweeteners, oils, fillers and flavor systems re-engineered. Barclays found in August 2023 that over half of British consumers had noticed products of lower quality at unreduced prices. Sylvain Charlebois at Dalhousie's Agri-Food Analytics Lab has tracked the pattern for years and, as of May 2026, reports the direction reversing: a return to simpler recipes and recognizable ingredients as consumers walk away. Hershey is reported to be moving back toward traditional chocolate recipes.

None of this happened because executives are careless. Cocoa prices reached historic highs. Input volatility was severe and the margin pressure was real. Reformulation is a legitimate tool and every company in the category faced the same arithmetic. The question is not whether they had a reason. It is whether they priced the reason correctly.

While the recipes were being quietly re-engineered, the store brand was going the other way.

Private brands are now a "structural commitment rather than cyclical opportunism."

Nik Modi, RBC Capital Markets, June 2026

The numbers say he is right. Store brand unit share reached a record 23.8 percent in the first half of 2026, with store brand units up 0.2 percent while national brand units fell 0.5 percent. And here is the line that should worry anyone running a brand: over the same period national brands added 2.2 percent in dollars while losing units. Revenue is being held up by price on a shrinking base of buyers, which is the precise condition that funds the switch away from you.

There is a body of work on what recessions do to store brand share, and it is not comforting. Lamey, Deleersnyder, Steenkamp and Dekimpe examined seven marketing-mix instruments across 106 US packaged goods categories over two decades. Private label share moves countercyclically, as everyone expects. What is less expected is that part of the gain during each contraction is permanent. Consumers switch faster and more completely in downturns than they switch back in recoveries. Each cycle leaves a residue.

The same research names the mechanism, and it is behavioral rather than structural. Retailers support their own brands during contractions and ease off in expansions. National brands do the opposite — they cut major new product introductions, advertising and promotional pressure exactly when the substitute is being pushed hardest. That procyclical conduct is associated with more pronounced share loss. One party leans in when it is difficult; the other leans out. Over enough cycles the outcome is not a mystery.

Which reframes the whole thing. Commoditization is not a state a brand gradually drifts into. It is a set of decisions taken on a schedule, in a specific quarter, by someone who could see the reason clearly and could not see the bill.

The uncomfortable part of the research is not that consumers are unforgiving. It is that they were willing to pay more all along, and would have thought better of you for asking. The recipe change bought a few points of margin and cost the one thing that made the price defensible in the first place.

Related from Sound Decisions: Distribution Doesn't Create Demand; It Captures It · Walmart Can End the Company · Nine Forecasts and No Evidence

The Rebate Closes Before the Shoot Starts. What $1.5 billion of Texas film money means for an agency.

Texas is spending about $1.5 billion on moving-image production through 2035, and commercials qualify. They collected about one percent of the last cycle, partly because the rates are low and partly because the window closes before most agencies think to ask.

"This is a film story. I run an ad agency." On the part everyone talks about, that is mostly right. In 2025 Texas committed about $1.5 billion to moving-image production through 2035, $300 million every two-year budget cycle, with grants of up to 31 percent of qualified in-state spending. Commercials qualify. But the headline rate is built for film and television, and the published record shows how little of the money reaches advertising.

That conclusion is right about the size and usually wrong about the reason. Commercials collect little partly because the rates are low, and partly because almost nobody asks in time.

The Texas Film Commission publishes every grant it pays, project by project, with the in-state spend and jobs behind it. Adding up the list for the two years ending August 31, 2025, six television projects took about 89 percent of the grant dollars. All thirty-nine commercials together took about 1 percent, $612,254, an effective return of about 6.25 percent on their qualified Texas spending.

The same list carries a jobs column. Those commercials reported 1,771 Texas jobs, about 15 percent of every job the program paid for that cycle, on roughly 3 percent of the spending. Per dollar of incentive, commercial production put more Texans to work than anything else the state funded, and collected the least.

The agency can be the applicant. For commercials, the Texas Administrative Code defines the applicant as the production company, the advertising agency, or the client. If the agency files and a production company spends the money, a production services agreement or chain of payments has to be shown. If the production company files, the grant is paid to the production company, and whether any of it reaches the client's budget depends on how the bid was written.

Several spots can be one project. The same code defines a commercial to include more than one advertisement created in a contiguous production period for the same client. The program's minimum is $100,000 of qualified Texas spending. Five spots at $30,000 each, shot in one block for one client, can be one qualifying project at $150,000 rather than five that each fall short. Whether a year of a client's work counts as one project or six is settled when the shoot calendar is set.

The rates are modest. Commercials earn 5 percent from $100,000 to $1 million of qualified spend and 10 percent above $1 million, plus possible additional awards of 1 to 2.5 percent. At least 60 percent of the production must be done in Texas, and at least 35 percent of combined paid crew and cast must be Texas residents.

At the bottom of the range the money is real and small; it will not move a shoot. It is not worth chasing job by job. It is worth installing once, because the second application costs almost nothing after the first.

The larger benefit needs no application at all. A decade of state money for long-form production is paying to build crews, stages and post-production capacity within driving distance of Texas sets. Those crews do not disband between seasons, and an agency quoting production on assumptions formed before the new money arrived may be quoting against a smaller local base than now exists.

The shares and effective rates are the author's arithmetic from the Film Commission's published line items for the 2024–25 cycle, before the larger fund began. The 2025–27 cycle may look different once its paid list is complete.

The jobs figures are self-reported by grantees and count jobs differently across project types; a day of commercial work and a season of television are not the same job.

Eligibility depends on details this piece does not cover: what counts as qualified spend, residency documentation, and the Film Commission's review of content and economic impact. Confirm with the Commission before relying on any grant.

Consolidating shoots to clear the floor must be a real production decision. Rescheduling work only to qualify can cost more than the grant is worth.

The state does not care who files. It cares when. Most of what agencies have left on the table was available for weeks, to anyone who asked before the shoot was scheduled.

Related from Sound Decisions: Houston Is Booming. Will Your Agency Boom Too? · How to Win Houston's Growing CPG Market · The Budget Line

The Memory Isn't Yours. You're asking to borrow it.

There is a comfortable theory that nostalgia marketing is what brands reach for when they have run out of ideas. The evidence does not support it. What the evidence does show is stranger, and harder to act on.

Someone in the meeting suggests bringing back the old logo, the old jingle, the packaging from 1987. Half the room lights up. The other half goes quiet, because they have heard the theory about what that kind of idea means, and nobody wants to be the brand that has run out of new things to say.

The theory goes like this. A business stops investing in product. Acquisition costs climb. Someone reaches for the archive. It works for a quarter because it is easier to borrow a feeling people already have than to earn a new one. Nostalgia, on this reading, is what decline looks like from the outside.

It is a satisfying theory. It lets you look at a retro campaign and diagnose a company. I went looking for the evidence to support it and found the opposite, twice over.

If nostalgia were a symptom of decay, the brands leaning hardest on it should be the ones in trouble. They are not. In 2023 Mattel grew market share in all three of its leading categories during an industry-wide decline, and collected an industry-leading fifteen Toy of the Year nominations for design innovation. Nostalgia and category-leading product development, same company, same year. Lego has been the world's most valuable toy brand for ten consecutive years, growing brand value another 6.5 percent to 7.9 billion dollars, while running one of the most deliberate heritage programs in consumer goods and one of the most serious design operations.

These are not companies that stopped investing in product and reached for the photo album. They are the reference cases for investing in product.

And then there is the question of who is actually driving this.

That is a fact about the audience, not about the companies serving it. A cohort that was seven years old in 1995 now has disposable income and no children in the room to justify the purchase. The wave was coming whether or not any brand had a strategy for it. Reading a demand-side shift as a supply-side confession gets the causality exactly backwards.

So the decline theory fails. But that does not make nostalgia a good bet, and the reason is the second thing I did not expect to find.

Every measurement in this field stops one step short of the cash register.

The academic literature is substantial and it measures purchase intention, brand attitude, engagement, and recall. A 2024 systematic review covering 139 indexed papers reports ambivalent findings and unresolved gaps rather than a settled effect. Brand valuation firms measure brand value and engagement. The widely repeated claim that people will pay ten to fifteen percent more for a nostalgic product traces to a research vendor's blog post with no primary source attached. The one genuinely hard behavioral finding, from the Journal of Consumer Research, is that nostalgia weakens a person's desire for money, which is a laboratory result about attitude toward cash rather than evidence of anything ringing up.

Then there is the largest natural experiment anyone could have asked for.

Attention is real and it is worth something. But attention is the input that every marketing function already knows how to buy, and the thing nostalgia is being sold as is a shortcut to demand. On that, the file is close to empty.

Here is the finding that reorganized how I think about this, and it comes from a 2019 paper in the Journal of the Association for Consumer Research.

People reach for retro-styled objects when their own sense of authenticity is under threat. The mechanism is self-continuity: the object links who they are now to who they were, and repairs the thread. The effect is real and it replicates. But it holds under one condition, and the condition is everything.

The object has to have past personal relevance to that person.

Read that carefully and the ownership question flips. The nostalgia does not belong to the brand. It belongs to the customer, it is doing a job for the customer that has nothing to do with the brand, and the brand is asking permission to be present while that job happens. Heritage is not a thing a company possesses. It is a thing a person grants.

Consumers are highly sensitive to what the literature calls "aesthetic exploitation."

Stephen Brown, on nostalgia deployed opportunistically

Which explains the failure mode precisely. Zhao and colleagues found that consumers accept nostalgic messaging when it lines up with a brand's actual history, and that a retro campaign from Coca-Cola or Nike reads as genuine where the same styling from a startup does not. Hartmann and Ostberg describe what happens when a revived brand turns out to be a simulation of the original rather than the original: skepticism, and a questioning of whether any of it was real. The trade term for borrowing someone else's visual past is flattering. The academic term is exploitation, and consumers appear to detect it.

None of this makes borrowed nostalgia impossible. It makes it a different product. If nobody in your audience has a memory of you, the vintage typeface is a style choice competing on aesthetics with every other style choice, and it should be judged that way rather than as an emotional shortcut. What it cannot do is the self-continuity work, because there is no thread to repair.

The comfortable theory was that nostalgia is what brands do when they have nothing left. The evidence says it is what customers do when they need to feel continuous with themselves, and that the strongest companies in a category are simply the ones most often invited to be there when it happens. That is a harder thing to buy. It is also a much better thing to have.

Related from Sound Decisions: Nine Forecasts and No Evidence · Hope Is Not a Strategy · Value Creation Beats Extraction

Nine Forecasts and No Evidence. Why product innovation actually fails.

The usual explanation is that too many people stand between a concept and a consumer. The data says something more uncomfortable: the number of people is not the problem. What they are being asked to do is.

Walk a new product through a large consumer packaged goods company and count the judgments it has to survive. A brand manager. A brand director. An innovation lead. Finance. Sales. A channel manager. A distributor. Somewhere in there, two or three rounds of commissioned research. Then, at the far end of all of it, a person in a store who has never heard of any of them and owes the project nothing.

The obvious conclusion is that the layers are the problem. Strip them out, get to the consumer faster, and the hit rate improves. It is a clean argument, it flatters everyone who has ever been slowed down by a gate review, and I believed a version of it myself.

It does not survive contact with the evidence. But the reason it fails is more useful than the argument was.

Almost every conversation about innovation failure opens with a statistic: 80, 90, sometimes 95 percent of new products fail. It gets attributed to Clayton Christensen, usually alongside a claim that 30,000 consumer products launch each year.

Christensen denied saying it. George Castellion and Stephen Markham asked him directly; he told them the rate was lower and observed that the field had never properly validated the figure. Their review found nineteen peer-reviewed studies between 1945 and 2004 putting market failure for launched products in the range of 30 to 49 percent. In consumer packaged goods the number runs worse than the cross-industry average, but nowhere near the folklore.

This matters more than a footnote. If you believe almost nothing works, you either stop trying or stop thinking, and both are expensive. The real rate is bad enough to take seriously and good enough to be worth getting right.

Here is the fact that reorganized how I think about this.

Four weeks. Whatever the gates were arguing about, the market resolves it almost immediately once the product is in front of someone with a basket and a budget. Everything expensive, slow and contested happens before that moment, in a period where nobody has the answer and everybody has an opinion.

And the opinions are not good. NielsenIQ ran 91 concepts for a large manufacturer that had been using an internal concept-pass system and found 62 percent of them had been sorted into the wrong category. The gates were not merely slow. They were wrong about which ideas were which, at a rate close to a coin toss. Worth naming the obvious: NielsenIQ sells the alternative system, so treat the precise figure as directional rather than settled.

Approval is also not the same thing as demand. Catalina studied the top-selling new food and beverage launches in the United States — the winners, not the strugglers — and found that of the shoppers who tried one in its first six months, just 11 percent were still buying it a year later. For the average new product in that set, 0.7 percent of shoppers accounted for 80 percent of volume. One shopper in 143 was deciding the fate of products that had cleared every internal review.

So: fewer layers, faster contact, better products. Except the evidence goes the other way, and it goes there four separate times.

Second, the denominator. If distance were the variable, the businesses with almost none of it should win. They do not. Research cited by James Richardson puts roughly 80 percent of premium food and beverage brands as never reaching a million dollars in retail sales. Coca-Cola's own venturing unit reported that around 3 percent of beverage brands reach ten million. Founders sit inches from their customers and fail at rates that would end careers inside a large company.

Third, structure. Robert Cooper's research puts the success rate for companies running ad hoc development at 24 percent, with structured, gated processes performing substantially better. More gates, better outcomes. That is the opposite of the prediction.

Fourth, and most damaging to the case for speed:

Accelerated development carries hidden costs, including "undertaking less innovative projects."

Robert G. Cooper, on compressing the idea-to-launch cycle

Compress the loop and you do not simply get the same ideas sooner. You get systematically smaller ones, because the ideas that survive a shortened process are the ones that needed the least explaining. Cooper also notes that the research on acceleration's benefits remains inconclusive — and in a separate study of 103 projects, that solid up-front work produced projects that were both more profitable and faster to market. Front-loading did not cost time. It bought it.

Put the four together and the picture inverts. Removing layers does not help. Adding structure does. Speed helps only sometimes, and it quietly narrows what you are willing to attempt.

But the original instinct was pointing at something real. The people approving the product are not the consumer, and they are not thinking like one. That part holds. The error is in what follows from it.

The problem is not that nine people stand between a concept and a consumer. It is that all nine are being asked to predict a consumer rather than observe one. Every gate is a forecast. Stack nine forecasts in sequence and you have not built rigor, you have built nine consecutive opportunities to be confidently wrong, with no contact with the ground at any point. The chain is not too long. It is unanchored.

That explains what distance cannot. It explains why founders with no layers still fail: they reach the market fast but cannot afford to learn from it twice, so their single contact with reality has to be right first time. It explains why structure beats its absence: gates ruling on evidence are genuinely useful, and gates ruling on opinion are theatre. And it explains why the compression units were shut down. They removed judgments without grounding the ones that remained, which changes how quickly you arrive at an unsupported decision but not whether it is unsupported.

Cooper's 692-project study lands the last piece. Speed to market improves outcomes, but market uncertainty moderates the effect — where uncertainty is low, speed barely matters. Read that as a statement about information rather than time. Going faster helps when you do not know something. When you already know, it buys you nothing at all.

The people in that chain are not the obstacle. Most of them are good at their jobs and being asked to do something nobody is good at, which is to know in advance what a stranger will do in a shop. The market will tell you in four weeks. The only real question is how much you spend guessing before you let it.

Related from Sound Decisions: Distribution Doesn't Create Demand; It Captures It · Hope Is Not a Strategy · Purely Elizabeth Took Eight Years to Raise $3 Million

Purely Elizabeth Took Eight Years to Raise $3 Million. That's Why She Was Still in the Room.

Ferrero announced on August 14 that it is buying Purely Elizabeth. Seventeen years after founding it, Elizabeth Stein is still the chief executive — and the reason traces back to a sequence of money decisions, not to the deal itself.

The first serious check usually arrives when the business needs it most. A funded competitor has just taken the endcap you wanted. The term sheet on the table would pay for the shelf, the trade spend and the team you can't afford. Saying no feels like choosing to stay small.

What it rarely feels like is a decision about the exit. It is one.

Almost every founder-exit story gets told backwards. The acquisition is announced, the press release names the price or declines to, and everyone reads the ending as though it were a single decision made in a conference room over a few weeks. It rarely is. By the time an acquirer is at the table, the founder's leverage has already been set — years earlier, by who they let onto the cap table and in what order. That is the decision most people skip past, because it happens quietly and it happens young.

Purely Elizabeth is a useful case precisely because the sequence is unusually legible. Elizabeth Stein founded it in Boulder in 2009 as a certified holistic nutrition counselor making ancient-grain granola. Ferrero — the company behind Nutella and Kinder — signed an agreement to acquire it on August 14, 2026. In between are two announced financing events, and the gap between the founding and the first of them is the whole story.

Eight years is a long time to run a consumer packaged goods business on its own revenue. It is slower than the category's own mythology recommends, and it costs something real: shelf space you don't buy, trade spend you can't fund, a marketing budget you don't have while a funded competitor does. Nobody should pretend that restraint is free.

What it buys is also real, and it is the thing that shows up seventeen years later. The founder who waits sells equity from a position of choice rather than necessity — and every subsequent round is negotiated against a business that already works, rather than against a runway that is ending.

Five years later, in January 2022, the company closed a $50 million Series B — co-led by SEMCAP's newly launched food and nutrition division, run by Haugen, the same executive who had led the 2017 investment from inside General Mills. Swander Pace Capital and Fresh Del Monte joined. Two announced rounds, roughly $53 million total, across seventeen years.

"We share the same vision and values for the brand."

Elizabeth Stein, on choosing a Series B partner, January 2022

Read that line as a founder rather than as a press release and it says something specific: she was in a position to apply that test. A founder raising against a closing runway does not get to weight vision and values above price and speed. That optionality was not created in 2022. It was created in the eight years when she did not raise at all.

There is a second, quieter fact worth naming. Her first institutional investor was General Mills' venture unit. General Mills is not the buyer. Nine years later the company went to Ferrero instead. Taking strategic money early bought her a knowledgeable board — it did not sign her exit over to the investor. Founders often assume corporate venture capital is a pre-negotiated ending. On this evidence, it is not.

So the lesson has to be taken from Act One, which is finished, rather than Act Two, which is not. What the sequence demonstrates is not that Stein got a good price — nobody outside the deal knows the price. It is that after seventeen years, two rounds and a sale to a multinational, she was still the person the acquirer had to keep. That is the outcome slow money protects.

The room at the end of a company's independent life is small, and the seats in it are assigned years before anybody sits down. Purely Elizabeth's founder kept hers. The eight years are why.

Related from Sound Decisions: Can Your Mission Survive Acquisition? · When the Money Changes Its Mind · Grüns Reached a Billion-Dollar Exit in Three Years

Hope Isn't a Strategy. Every Strategy Starts as One.

A plan can be approved, sized, scheduled and fully staffed and still be hope wearing a plan's costume. The difference is three things you can write down, and most companies never write the third one.

There is a particular kind of relief in the room when a big move gets announced. The reorganization. The spin-off. The rebrand. The new platform. Everyone has been carrying the problem for so long that the announcement itself feels like progress, and for a few weeks it genuinely is. Nobody asks the awkward question, because the move is large, and it is happening, and surely something that large will do something.

Hope is not a strategy is one of the oldest lines in management. It is also, as normally used, close to useless, because it gets aimed at the wrong target. It is used to scold optimism. Optimism was never the problem. The problem is that hope does not usually show up looking hopeful. It shows up looking like a plan: board-approved, costed out, with a timeline and a project team and a name.

The test is not whether a move is specified. It is whether it reaches the cause.

A split, then a reversal

In September 2025, Kraft Heinz announced it would separate into two independent public companies. One would hold the sauces and spreads, the part that was still working. The other would hold the grocery staples. The stated reasoning was the reasoning these announcements always carry: two more focused organizations, less complexity, brands better able to compete. The board approved it. A new chief executive was hired specifically to run it through, arriving on the first of January.

Six weeks later, on February 11, 2026, he stopped it.

What makes this worth studying is not the reversal. Companies reverse things all the time, usually while insisting they haven't. What makes it worth studying is that he said out loud what the split had never addressed.

"We have historically underinvested in our brands."

Steve Cahillane, CEO, Kraft Heinz · Q4 earnings call, February 11, 2026

He went on to tie that underinvestment directly to persistent share loss across the previous decade. He did not say the split had been a mistake. He said the board had reached a reasonable conclusion with what it had. That is the honest version, and it is the one worth sitting with, because the split was not stupid. It was structural. And the problem was not structural.

Reorganizing which company owns which brand does not change whether anyone wants the food.

What the numbers had been saying the whole time

The 2025 results carried the diagnosis in a single line, for anyone reading the composition rather than the headline. Organic sales fell 3.4 percent for the year. Inside that number, price was up 0.7 points and volume and mix were down 4.1 points.

Read plainly: the company was charging more and selling less. Fewer units, leaving the shelf, at higher prices. That is not a complexity problem and no amount of corporate architecture touches it. It is the signature of a business capturing value faster than it creates any, and it can hold up the revenue line for years while the thing underneath quietly empties out. The near ten billion dollars of brand write-downs booked that year were the accounting system catching up to what the volume line had been saying for a decade.

Three things that turn hope into a path

The same day the split was paused, the company announced it would put six hundred million dollars into marketing, sales, research and development, product quality and price. On its own, that is just a different big move, and big moves are cheap to announce. What separates it is not the size. It is that it carried three specifications the split never did.

The second row is the rare one, and it is the one I would ask you to steal.

They published the bill before they published the benefit. Guidance issued that morning said full-year organic sales would fall, that operating income would drop by double digits precisely because of the investment, and that earnings would land below what analysts were expecting. A company announcing a turnaround told the market that its numbers would get worse first, and put a range on how much worse.

That is what a costed path looks like, and almost nobody does it, because publishing the cost is the part that can make you look wrong in public. It is much safer to announce the benefit and let the cost arrive quietly, one quarter at a time, where it can be attributed to the weather.

Six months in, the test is running

By the second quarter of 2026, reported on August 5, the decline had narrowed considerably. Organic sales were down 1.3 percent against down 4.2 percent two quarters earlier. The company raised its full-year sales outlook, raised its earnings outlook, and then raised the investment itself by another hundred million. Only about a third of the original money had been spent.

And it still might not work

This is the part where a piece like this normally lands the plane, and it shouldn't.

Organic sales are still negative. The second quarter carried a further seven point four billion dollars of non-cash impairment, on top of the nine point three billion the year before. Third-quarter guidance is worse than the second quarter, not better. And the company is still taking price while it invests, which is the same lever that helped hollow out the volume line in the first place.

So the honest reading is narrow and it is specific: the path is working on the leading indicator and has not yet worked on the lagging one. That is a real thing to know and it is not the same as a recovery. Anyone telling you in August 2026 that this turnaround has succeeded is doing exactly what the split did, which is offering a conclusion the evidence has not reached.

Your version is smaller and it is the same

Almost nobody reading this is allocating six hundred million dollars. It does not matter. The shape is identical at every scale, and it is probably sitting in your plan right now: the rebrand, the new sales hire, the platform migration, the second product line, the agency review, the market you are about to enter.

Ask the first question first. Does the move reach the cause, or does it rearrange the furniture around the cause? A rebrand does not fix a product people have stopped enjoying. A new salesperson does not fix an offer nobody wants at that price. Restructuring the team does not fix work the client no longer values. These are all real moves, fully specifiable, and none of them touch the thing that is actually happening.

Then apply the three tests, and expect to fail the second one. Naming the move is easy. Setting a test is uncomfortable but doable. Writing down what it will cost you, in advance, where someone else can see it, is the part almost everyone skips, and skipping it is precisely what lets a plan stay hopeful indefinitely. If the bill never got written down, the plan can never be seen to be expensive, and if it can never be seen to be expensive, it never has to be judged.

Hope is not the enemy here. Hope is the raw material. Every strategy worth running started as somebody hoping something would work. It stops being hope, and starts being a path, at the exact moment you name the move, price it out loud, and say what you will accept as evidence it isn't working.

Creative Is Now a Line Item. Three holding companies priced it three ways.

Between July 16 and August 6, the three largest agency holding companies each reported what creative work is worth inside their business. One is moving capital away from it, one reported it shrinking, one is growing it slower than media. What they all did first is the finding.

If you run an independent agency, holding-company earnings season usually reads like weather in another country. Big numbers, distant companies, headlines about restructuring that have little to do with the brief on your desk. You skim them and go back to work, because the question you actually live with, what a client will pay for your creative thinking, never seems to be in them.

This summer it was. Publicis reported on July 16. Omnicom on July 28. WPP on August 6. Each was covered on its own, as a verdict on one company. Read side by side, the three filings answer a question independent agencies rarely get a straight answer to: what is creative work worth to the companies that own the most of it?

They disagree. One is selling advertising businesses. One reported its creative agencies shrinking while its production unit grew. One reported creative growing, more slowly than the media business beside it. Same half-year, overlapping clients, the same technology, and three different answers from three leadership teams that are not short of information. That disagreement is useful to anyone smaller. A settled question leaves nothing to claim. An open one is a position still available.

Omnicom is moving capital away from it. Its quarter was strong: core operations revenue of $6.0 billion, organic growth of 6.1 percent, margins up. Underneath, $567.5 million of revenue in the quarter came from businesses disposed of or held for sale. Chief executive John Wren described what the combined company is for in one line: “We are built for an era where speed, integration, and scale matter most.”

WPP reported the split in its own numbers. Revenue less pass-through costs fell 4.7 percent like-for-like in the first half. Inside that, WPP Creative declined 4.9 percent while WPP Production grew 1.6 percent. Average headcount fell from about 106,000 to 97,000 year on year. Read carefully, the creative figure is less dramatic than it first looks: it includes WPP's enterprise-solutions unit, WPP attributes the decline to lower client spending, and it reports its brand and design agencies still growing.

Publicis is growing it, and even there it is the slower line. Connected Media, 62 percent of net revenue, grew at a high single-digit rate in the second quarter. Intelligent Creativity, 25 percent, grew at a low single-digit rate. The group raised its full-year guidance. Creative grew; media grew faster. What Publicis did was not protect creative so much as package it, selling it inside a connected offer with data and media and naming it for what it does.

Omnicom calls it Advertising. WPP calls it Creative, one of four operating units. Publicis calls it Intelligent Creativity. Different words, the same move: in all three companies, the work is now a unit with its own revenue line, its own growth rate, and a quarterly comparison against units that scale better. That happened before any of the three decisions, and it is what made them possible. A unit can be sold, shrunk or renamed. Something that organizes the whole firm cannot.

An independent agency has the opposite structure, and it is an advantage. Creative is not a line on its profit and loss; it is the whole business, and nobody can reorganize it away. The exposure is the other side of the same fact. The 4A's and the ANA reported in June 2024 that only 10 percent of agencies, and 5 percent of clients, have a formal definition of the value an agency delivers. Publicis wrote its definition into how it sells. Most independents never have.

One reporting cycle, three companies. Omnicom's figures are reshaped by its acquisition of Interpublic, which closed in November 2025; WPP is mid-restructuring; Publicis is on a different fiscal rhythm. Third-quarter results arrive in late October and may change the picture.

The discipline names are the companies' own, and they are marketing labels as much as accounting categories. WPP's creative line includes a consulting-style unit; Omnicom's advertising line is not the same set of businesses as Publicis's creative practice.

The strongest argument against all of this is Publicis itself. If packaging creative inside a connected offer is what keeps it growing, a holding company can build that package faster than an independent, because it already owns the data. What can be said is narrower: a package built on data belongs to whoever owns the data, and one built on judgment belongs to whoever makes the judgment.

The three largest owners of creative work reached three different answers about what it is worth, in the same three weeks. That is not a reason for anxiety. It is the stretch of time in which the definition is still open for an independent firm to write.

Related from Sound Decisions: They’ll Pay for the Work. They Won’t Pay for the Thinking. · You Don’t Need More Clients. You Need Longer Ones. · Your Best Clients Are the Ones You Undercharged

Leaving isn't the skill — leaving with something is

Every year, capable people exit structures that were capturing most of the value they created, and most of them are never heard from again. The difference is rarely courage. It is what they were carrying on the way out.

There is a conversation that happens in every independent firm, usually late, usually after a project that went well for everyone except the people who did it. Some version of: we are the reason this worked, and we are not the ones capturing it. Should we be doing this on our own?

The question is almost always framed as a matter of nerve. It isn't. It is a matter of inventory.

I want to use a case that isolates the variables unusually well, because the same person made both decisions in public and the results are a matter of record.

Generating value, positioned to capture almost none of it

The singer RAYE was signed to a major label at seventeen on a four-album deal. Across roughly seven years she released one nine-track mini-album. In 2021 she said publicly that she had albums of her own material sitting unreleased while she gave songs away to A-list artists, waiting for permission to put out a record of her own.

Look at the structure rather than the story, because the structure explains everyone's behavior including the people who look like villains.

A traditional recording contract transfers the master recordings to the label, in perpetuity. The artist receives a minority royalty, and receives nothing until the advance and a long list of recoverable costs have been earned back. The label's return comes from owning an asset forever. The artist's return comes from a small share of income that begins only after a debt clears.

Those positions diverge most sharply on the exact question they were fighting about. For the label, an album is a large capital-intensive bet. For the artist, an album is the entire career. A label with an artist already generating safe revenue as a featured vocalist has very little reason to take the bet.

Nobody has to behave badly for that to produce seven years of silence. The incentives produce it on their own. Which is worth holding onto, because it means the fix is never a better relationship with your counterparty. It is a different structure.

She did not leave empty-handed

I want to be careful not to make this redemptive. Seven years of unreleased work is a real loss that nothing recovers, and the eventual success does not retroactively justify the arrangement that caused it.

The point is narrower. People who go independent successfully almost always go independent late — carrying capability, relationships and proof they accumulated inside a structure that was capturing most of the money while they accumulated them. The question is not whether you are unhappy enough. It is what is portable, and whether you have leverage you are actually willing to use.

She didn't bypass the majors. She changed the contract.

Eleven months after leaving, she signed with a company owned by Sony Music and distributed through Sony's distribution arm.

Read that again, because it dismantles the version of this story that gets shared. She left one major-label group and signed with another. She did not escape the system. She changed her position inside it.

What changed was the contract type. Under the old deal the label owned the masters, made the release decisions, and paid a minority royalty after recoupment. Under the new one she owns her master recordings, makes her own release decisions, and pays a fee for distribution infrastructure it would be absurd to rebuild from scratch.

Same scale of counterparty. Opposite direction of ownership.

This is the most transferable idea in the case. The romantic version of independence says get out from under the large players and own everything yourself — which is usually a route to being small permanently, because the infrastructure genuinely does something and rebuilding it costs years you don't have.

The useful version says work with whoever has the infrastructure, and fight about ownership rather than about association. Ask what the contract does with the rights, not how big the logo is.

This is survivorship evidence

Thousands of artists leave label deals every year and disappear. She left carrying an unusual set of assets, and the independence did not cause the outcome — it removed a constraint on an outcome the preceding decade had already made possible. If the lesson you take is “go independent,” you have taken the wrong one, and it is the expensive kind of wrong.

The timing helped too. Independent distribution economics in 2022 were dramatically better than a decade earlier, because distribution had become purchasable rather than gated. Some of what looks like judgment was a door that happened to be open.

What survives both caveats is the structural point, and it is the one worth acting on. Most independent firms negotiate hard on price and softly on rights, because price is concrete this quarter and rights feel abstract. That asymmetry is why so many owners discover, years in, that the thing they built belongs to somebody else — settled in language nobody has reread since it was signed.

Related from Sound Decisions: The Ground Nobody Was Defending · Value Creation Is a Business Model, Not a Moral Posture · Additions Are Free

The ground nobody was defending

The challenger's real starting position isn't unknown. It's unattributed — generating value that gets credited to someone else. What comes next is not a better product. It's a piece of ground the incumbent has already decided isn't worth holding.

You have probably seen your own work presented by someone else. The strategy that went up on the screen under the name of the firm that subcontracted you. The system your client's in-house team now maintains as their own. The pitch you built that won work you didn't get to do. Everyone in the room liked it. Nobody in the room knew it was yours.

That is a more common position than the one most challenger advice is written for. The usual picture is a scrappy outsider with a better product and no audience. The more frequent reality is a capable operator generating real value as an input into somebody else's branded output, and the difference matters, because it leads to a different strategy.

You are not unknown. You are unattributed. Those are different problems with different solutions.

Good enough to be commissioned, not positioned to be chosen

The singer RAYE has a phrase for the decade she spent in this position. She called herself a rent-a-verse — people knew her songs, but they didn't know her.

What she did next is the interesting part, because it did not involve competing with anyone.

Take ground that is empty for structural reasons

British pop has a dominant occupant in the emotionally serious, piano-and-strings register — a position held at the highest level for over a decade, with catalogue depth and global recognition. Competing there means being measured against that standard, on that standard's terms, by an audience with a fixed reference point. Challengers who try it get described forever as the new version of the incumbent. That is not a position. It is a comparison you lose slowly.

She went somewhere else. Her breakout single is built on big band brass, swing feel, gospel-inflected backing vocals and a rapid, almost percussive delivery. Theatrical, funny, and deliberately archaic in its instrumentation.

Nobody in mainstream pop currently occupies that ground — not because it's bad, but because it's expensive, requires live players, doesn't reduce to a production template, and demands a vocalist who can genuinely swing. The category was empty for structural reasons, and those same reasons are what made it defensible once she was standing on it.

Choose the axis your competitors won't follow you down. Not the axis where you're marginally better — the axis where following you would require them to rebuild something they have deliberately optimized away.

Change the texture, keep the format

Here is where most differentiation strategies quietly fail, so it's worth being precise about what she didn't do.

The single runs three minutes and sixteen seconds. It was debuted live at a festival before release, engineered for a crowd to sing back, released and promoted conventionally, and reached number one on the UK singles chart fourteen weeks later — climbing forty-seven places in the week it topped.

She changed the sonic signature and kept every commercial convention of the category she was entering: the runtime discipline, the hook placement, the release strategy, the testing.

Differentiation that abandons a category's functional requirements isn't differentiation. It's exit. Plenty of independent firms have positioned themselves out of the market by mistaking “we work differently” for a proposition, when what they have actually done is become harder to buy. Be unmistakably distinct on the dimension buyers notice. Be entirely conventional on the dimensions that make you easy to hire — scoping, contracting, billing, delivery predictability.

Borrow credentials, and make the borrowing expensive

Standing in a new category is a claim, and claims made by challengers are discounted by default. You can spend years earning standing alone, or you can get third parties with existing authority to vouch — which is faster, and which a competitor cannot replicate if they can't get the same institutions to show up.

For an independent firm the equivalent is not a partner-page logo. It is the joint research with a university that took nine months. The industry body you contributed real work to for two years before anyone noticed. The stage you earned by having something to say.

You've arrived when value flows the other way

In March 2026 she released a track featuring Hans Zimmer. Read as a prestige play, that looks like borrowed credentials in their purest form: pop artist enlists legendary film composer, absorbs his gravitas.

Then look at the charts. The song gave Zimmer the biggest chart success of his career in the United Kingdom, delivering new career highs across multiple charts — for a composer whose previous best on the singles downloads chart was a novelty track from an animated film. Meanwhile it did not rank among her own biggest releases. Her lead single had already been a number one.

She did not borrow his prestige. She gave him reach.

That reversal is the clearest available signal that a challenger has finished being a challenger. There's a test in it any firm can run on itself: when you enter a partnership, a co-authored piece of work, a joint venture — who is getting more out of it? For a long time the honest answer is “us, and we should be grateful.” The day it flips is the day the position is real, and most people notice late, because they're still negotiating like the junior party.

The category does not stay won

At the 2026 BRIT Awards she was nominated in two categories and won neither. A younger artist, working in a different register on undefended ground of her own, took four including Artist of the Year — two years after RAYE's own record-breaking six-win night, which is still the most any artist has won in a single ceremony.

Awards are not revenue, one ceremony is not a trend, and she released a number one album a month later. By any commercial measure she is in the strongest position of her career. And one artist's career is an analogy from music, not evidence about firms; it shows the moves clearly, not how often they work.

But the structural point stands, and it is the most important line here for anyone running a challenger business. She is now the incumbent, and somebody else is running the playbook at her.

That is not a failure of the strategy. It is how categories work. Positions are rented, not owned. The three moves are not a ladder you climb and then stand on — they're a cycle, and the moment you start defending rather than moving is the moment you become the fixed reference point somebody else differentiates against.

Which means the operative question for a challenger who has just arrived is not how to protect this. It is what the next piece of undefended ground is, and whether you are still willing to go stand on it before anyone has given you permission.

Related from Sound Decisions: Leaving Isn't the Skill — Leaving With Something Is · Additions Are Free · Value Creation Is a Business Model, Not a Moral Posture

Additions are free — which is exactly why they don't work

A brand the market has already rejected cannot be argued into acceptance. More weight behind an unbelieved claim makes the gap more visible, not less. What moves a position is what you can be seen to have paid.

Every marketing leader who has tried to move a brand upmarket knows the meeting. The proposition is agreed, the new positioning is signed off, the pack redesign is beautiful, the price architecture is modeled, and the plan for making it real is a larger budget behind a better-articulated claim.

Then it doesn't work, and the diagnosis is that the message needs more time or more weight. So it gets more of both.

I want to walk through a case where that exact loop ran for four years in public, with the results measured quarterly, before someone finally read it correctly. The company is WWE and the asset is a performer, but nothing about the mechanism is specific to either.

Four years of maximum support

In January 2015, a performer the company had spent two years preparing won the biggest match of its calendar. To make certain the moment landed, they brought out one of the most beloved figures in the company's history to stand beside him and raise his hand.

The crowd booed harder.

What followed is the standard playbook for establishing a premium asset. More screen time. Bigger matches. The main event of the biggest show of the year, four years running. The world championship, then two more titles. More borrowed credibility from more legendary figures. When audiences rejected it, the company increased the support.

Now look at what all of those interventions have in common.

Every one of them was an addition, and additions are free. Giving a performer more airtime costs the company nothing it wasn't already spending. Handing him a championship is a bookkeeping entry. Standing a legend beside him borrows credibility rather than building any. Each move was an assertion that this man belonged at the top — and assertions cost nothing to make, which is precisely why audiences discount them.

The audience had correctly identified a gap. The claim was that he had earned the top position. The perceived reality was that he had been assigned it. Every new accolade widened the gap rather than closing it, because each one was further evidence that the position was being conferred rather than won.

They stopped arguing and started paying

The turn began in 2020, and it started from a different reading of the same evidence: the audience is not wrong. For five years they had been saying this man's position looked conferred, that he acted entitled to something he was handed, that he expected deference he had not earned. The company had treated that as a misunderstanding to correct. What if it was an accurate product description?

So they made it the proposition. He returned demanding acknowledgment, imperious, insecure beneath it, certain of his right to the top spot. A figure who demands unearned deference cannot be criticised for demanding unearned deference. The objection was absorbed rather than rebutted.

But adopting the audience's read was still only a claim, and the previous five years had established that claims did not work here. What made this one credible was what the company and the performer gave up.

The generic tactical vest — the same look worn by a dozen performers across the industry — came off, leaving a body and a heritage tattoo nobody else has. The ring style slowed down, trading immediate crowd reaction for long-term credibility, which is a bet that only pays after years. He spoke less, and handed a large share of his own promotional time to a manager, sacrificing screen presence in a business where screen presence is the currency. He surrendered the hero role — the commercially conventional slot, the one with the children's merchandise and the uncomplicated sponsor conversations.

And in 2022 he signed a contract that substantially reduced his schedule, moving from a full touring calendar to selected major events. His own explanation is the cleanest statement of the principle I have heard from an operator: it makes everything feel more special if you're not as attainable or available.

A signal is credible in proportion to what it would cost to fake

This is the whole mechanism, and it is why the two attempts produced opposite results despite pursuing the same objective.

Cheap talk gets discounted. Everybody can say they are premium. Nobody can cheaply reproduce a change that took real money, real risk, or real forfeiture — which is why those changes function as evidence and advertising weight does not.

The first attempt was made entirely of things that cost nothing to give. The second was made of subtractions and non-copyable investments: exposure surrendered, availability cut, a commercially safer role abandoned, immediate reward traded for slow credibility.

There is a second-order effect worth naming, because it is where most premium strategies quietly fail. Availability and premium pricing are inversely related. The instinct when something works is to increase availability — more doors, more SKUs, more formats, more occasions. It feels like maximising return on a proven asset. It is usually premium erosion executed one defensible decision at a time, because every increment removes a little of the scarcity that justified the price.

And restricting availability is not free either. Pulling the top asset out of weekly circulation left a hole in the product; the company's eventual answer was to introduce a second championship to fill it. If you protect a hero SKU by limiting distribution, something has to fill the shelf gap — a second line that costs money and risks cannibalisation, or accepted absence and the share loss that follows. There is no version where the scarcity benefit arrives for free.

Costly signals build equity. Price spends it.

The company reported its second quarter on August 3, 2026: revenue up 18%, adjusted EBITDA up 23%, full-year guidance raised. Underneath that, live attendance through the first half of 2026 was down 23.5% year on year, the marquee event was down 15–17%, and the summer event drew roughly 35,700 and 32,300 across two nights against 53,000 and 60,000 the year before — after tickets that opened between $200 and $3,000 were cut to as low as $25 to move inventory. Average ticket prices for regular North American shows had risen from about $75 in 2024 to about $118 in 2025. Live event revenue fell roughly $33.7 million year on year in the quarter while the segment still posted a 12% gain on media rights.

That shape will be familiar: the price ladder taken one rung too far. Volume falls, price covers the revenue line for a while, other parts of the business mask it in the consolidated number, and the demand signal reaches someone with authority several periods late.

It is also the exact inverse of the mechanism that built the position. Costly signals are investments in something hard to fake. Price increases extract from the equity those investments created without adding to it. You can run extraction for a while because the equity is there — and every increment spends down the thing that made the price defensible in the first place.

In fairness: live entertainment pricing has outrun inflation broadly since 2023, 2025 was an extraordinary comparison, and two quarters is not a trend. But the direction is legible, and it is the ordinary failure rather than a dramatic one. Not a strategic error. Not a creative collapse. Price, pushed one rung too far, in a period when the headline numbers looked good enough that nobody had to ask.

Which is the part worth carrying. The repositioning is the hard, visible, celebrated work. Holding it is the quiet work — and it fails in the years when everything looks fine.

Related from Sound Decisions: Your Best Year on Paper Can Be the Year the Brand Starts Dying · The Marketing Job Is Derived Power · Value Creation Is a Business Model, Not a Moral Posture · Leaving Isn't the Skill · The Ground Nobody Was Defending

The marketing job is derived power — and that's the design, not the flaw

You don't own the product, the roadmap, or ultimately the customer. Your entire function is to increase what the market will pay for something that isn't yours. The people who do it best stopped resenting that a long time ago.

Most people who take a senior marketing role want a version of it that doesn't exist. They want the authority of a founder with the mandate of a functional lead — to set the direction, own the thesis, and have the name on the outcome. Then the first budget cycle arrives, product ships something they weren't consulted on, and sales renames the category in a customer meeting, and the job reveals its actual shape.

I want to make the case that the shape is correct, and that treating it as a defect is the single most expensive mistake available in the role. The clearest working example I know comes from an industry most executives would never study, which is exactly why it's useful — the craft is visible because nothing else about the setting is familiar enough to distract you.

Six lines, and none of them works alone

Paul Heyman has spent four decades in professional wrestling as a manager, promoter, writer, and executive. He is best known for short pieces of speech that introduce someone else. At SummerSlam in July 2022, in front of roughly 48,000 people at Nissan Stadium in Nashville, he introduced Roman Reigns ahead of a title defense against Brock Lesnar. The introduction ran about six lines, and each line is doing a different job.

It opens on threat — “the most violent, the most diabolical, the most dangerous” — three adjectives escalating, setting category and register before any claim is made. Then proof by competitor: “a beast slayer,” then “a conqueror of conquerors.” Not a claim about the product but a claim about what the product has beaten, which is the strongest evidence available because the audience watched it happen. Then the category claim, staked openly with no hedge: “the goat of all goats.”

Then the line that should stop a marketing leader. In the middle of a piece of entertainment, sold to a paying audience, sits an explicit commercial assertion — “the ultimate needle mover.” Not the best performer. The one who moves the numbers. Almost nobody says the quiet commercial part out loud to the customer. It sits fourth.

Then the owned names — “the head of the table,” “the tribal chief” — rather than “champion,” which is a category term any competitor could use verbatim. And then a close on feeling instead of argument: “in god mode himself.”

Threat, proof, category, commercial claim, owned name, emotion. Read as marketing the sequence is unusually disciplined — most brand teams could not assemble those six moves in a quarter with a budget.

But the structure is not the interesting part. None of those lines is doing its own work. Each is a withdrawal from an account that had already been filled. Beast slayer only lands because the audience watched the slaying. Tribal chief isn’t a description — it’s a term that took years of accumulated story to mean anything, and by the time it’s spoken it arrives carrying all of it.

That’s why it can be six lines instead of six paragraphs. He isn’t building a position. He is spending one.

This is what brand equity actually is at the moment of use, and it is nothing like the way it usually gets discussed. Not a logo, not a guideline document, not a set of adjectives in a deck. It is the ability to say a short phrase and have an entire history arrive with it. Most organizations never reach that state. They are still explaining themselves in full sentences in year six, because nothing they have said has been held long enough to compound into a name.

He ran his own company once. It went bankrupt.

Before any of the work he's known for, Heyman owned and ran a wrestling promotion. He set the pricing. He managed a distribution stack across syndication, cable, pay-per-view, and home video. He made payroll weekly. He did all of it while two vastly larger competitors raided his roster continuously. This was not a creative with no commercial nerve — it was full P&L ownership under sustained competitive attack.

He was also, by any reasonable measure, right about the market. He identified an audience the larger companies were underserving, built a product for it, developed talent the entire industry later drew from, and was correct enough that his competitors copied his approach and built their most profitable era on it.

Set the ethics of that period aside — they are contested, they are not mine to adjudicate, and people who trusted him lost money they were owed. The structural fact is the one a marketing leader needs: he had extraordinary product-market fit and no revenue floor. No guaranteed television money. No capital reserve. No income that arrived regardless of last week's performance. His competitors had all three.

He then spent twenty-five years working inside a company that had the floor, and produced the most commercially successful long-form brand story of the modern era in his category.

Same person. Same instincts. Opposite financial outcome. The variable was the floor.

If you are a marketing leader with genuine conviction and a business that won't fund it, this is your case study, and it does not end the way you want. Vision without a floor doesn't get vindicated. It gets copied — by organizations that have one — after you run out of money.

Refusing to reuse the thing that worked

Most positioning failures in commercial marketing are this exact failure, and they are hard to catch because they feel responsible. An agency wins with an approach and applies it to the next client. A company finds a message that worked for one product and extends it across the line. A leader builds a reputation for a certain kind of work and then does that work regardless of what the situation needs. Every one of those is reuse of a proven method in a context where the proof doesn't transfer — and it gets defended as evidence-based, which is why nobody stops it.

The corollary is the vocabulary problem. The language built for the second client was proprietary — terms invented for one asset that a competitor could not borrow without the borrowing being obvious. Most brands never build any. They describe themselves in category terms any competitor could use verbatim, and then wonder why six years of consistent spend hasn't produced a position.

Amplification is the discipline, not the consolation

Look at what keeps happening around this man. A performer he managed early became a permanent brand asset built around a single memorable number. A phrase describing association with him entered the audience's vocabulary and stayed there — the affiliation itself became worth something. A performer the audience had rejected for five years was rebuilt into the center of a record-setting period. An opponent, not even a client, was repositioned by a single ten-minute exchange that observers credit with rescuing the biggest event of that year.

He does not accrue value to himself. Every label he has used — manager, advocate, counsel, wise man — is deliberately subordinate. His entire professional identity is built around raising the value of an asset that isn't his.

That is the job description, not a smaller version of a better one. The product isn't yours. The company isn't yours. The customer relationship, in the end, isn't yours either. Your function is to take a thing you don't own and increase what the market will pay for it, using positioning, language, talent, timing, and a read on what people are about to want. Done well it is among the highest-leverage functions in any company. Done resentfully it produces a marketing leader who spends the role trying to prove they should have been given a different one.

Where the analogy stops

Two things this case doesn't carry. First, the period where he held full commercial accountability is the period that failed; the work everyone cites as his best was done as a specialist inside someone else's P&L, where he owned positioning and talent but not pricing, distribution, or the bottom number. The complete job and the peak work are not the same decade of his career, and anyone using him as a model should know which version they're copying. Second, attribution is genuinely difficult — the results of the last five years belong to a system with many authors, not to one person. Read this as a case study in a craft, not a measurement of an individual's output.

What survives both caveats is the mechanism. Equity compounds into names you can spend later. Proven formulas carry associations that don't transfer. Language you own is the only kind that defends itself. And the marketing job is derived power by construction — which is a description of how it works, not a verdict on how much it matters.

Related from Sound Decisions: Your Best Year on Paper Can Be the Year the Brand Starts Dying · The Spend You Inherited Is Two Different Problems · Value Creation Is a Business Model, Not a Moral Posture · Additions Are Free — Which Is Exactly Why They Don't Work · Leaving Isn't the Skill · The Ground Nobody Was Defending

Your best year on paper can be the year the brand starts dying

Contracted revenue keeps arriving long after the relationship that earned it has begun to fail. That gap — between what the income statement says and what the audience is telling you — is where brands quietly go broke.

There is a specific kind of comfort available to a marketing leader whose revenue is mostly contracted. The renewals are booked. The distribution agreements run for years. The quarterly number lands where it was supposed to land, and the deck writes itself. Nothing in the reporting suggests urgency, because nothing in the reporting is designed to detect the thing that is actually going wrong.

I want to take you through a case where that gap ran for five years, in public, at a company most executives would never think to study — and where the reckoning, when it came, was survivable only because someone had bought enough time. The company is WWE. The lesson has nothing to do with wrestling.

Record financials, collapsing audience

In 2019, WWE was, by the measure most boards use, in the strongest position in its history. New television rights agreements for its two flagship shows had transformed the income statement — better than the 1980s boom, better than the company's celebrated late-90s peak. An operator reading the P&L would have concluded the brand had never been healthier.

Both things were true at once. Every audience-facing indicator was in decline. Every financial indicator was at a record. And the financial indicators were at a record because of contracts signed before the decline, which paid the same regardless of whether last week's product was any good.

This is the part worth sitting with, because it is the version of the problem most marketing leaders will actually face. The company was not ignoring bad news. It was reading good news, from instruments that were structurally incapable of showing the bad news for another two to three years. The dashboard was green. The foundation was cracking. Nobody was lying to anybody.

A revenue floor funds patience — or denial

The contracted revenue in that business is what I'd call a floor: money that arrives on schedule, largely independent of short-term performance. Every marketing leader wants one. Renewals, multi-year distribution, retainers, licensing, committed shelf — whatever the category equivalent is.

A floor is genuinely valuable, and here is the honest version of why. It is the only thing that makes long-horizon brand building affordable. You cannot run a three-year positioning strategy on revenue that depends on this month's conversion rate; the moment the quarter wobbles, the strategy gets cut, and it gets cut by reasonable people making reasonable decisions. Patience is not a virtue you summon in a planning session. It is a position you can afford or you can't.

But the same instrument does something else, and almost nobody plans for it. A floor also buys the ability to postpone a reckoning. It keeps paying while the relationship underneath it decays, which means it removes the pressure that would otherwise force you to look. From inside the organization, funded patience and funded denial produce identical reports. Both look like stability. That is not a metaphor — it is a measurement problem with a specific shape, and it is solvable only by watching instruments the contracts don't touch.

They stopped arguing and used what they heard

In 2020 the company changed approach, and the change is more instructive than a simple reversal would have been. They did not replace the performer. They took the exact thing audiences had been rejecting — the sense that his position had been conferred rather than earned — and made it the explicit content of the character. He became a figure who demanded deference, was entitled about it, and was visibly insecure underneath. The audience had been telling them something accurate for five years. The company finally built the product out of it.

What followed is the part a marketing leader should care about. Between 2020 and 2024, the company set records across ticket gates, attendance, merchandise, and viewership. In 2023 it drew higher total and average attendance than it had in 2019 — while running roughly ninety fewer events. That comparison is the cleanest evidence in the whole story, because it is independent of pricing: more people, fewer shows.

Then, in January 2024, the company signed a ten-year streaming agreement for its flagship reported at more than $5 billion. The floor got dramatically larger — after the relationship was repaired, not before. The sequence matters. Demand recovery preceded the contract; the contract did not cause the recovery.

Did proximity to the hit make everything else more valuable?

One more mechanism, because it is the transferable part and it applies directly to a portfolio.

The story built during that recovery was designed so that being close to the central figure made you more valuable — even in defeat. Supporting performers who lost their biggest matches emerged from those matches as bigger commercial assets than they went in. One went from a competent ensemble player to a solo headliner with his own merchandise line. Another, who lost the most emotionally charged match of the period in his hometown, became a permanent main-event fixture off the back of it.

Contrast that with the far more common pattern, in which a dominant asset absorbs value from everything around it. The flagship product takes the budget, the senior attention, and the best people. Adjacent lines get starved. It looks efficient — right up until the flagship softens and you discover there is nothing else, because you spent five years preventing anything else from developing.

When WWE's central figure stepped back from the top position in 2024, the business did not fall off a cliff. It had spent four years manufacturing successors inside the story. That is the actual test of whether you built a value-creating system or an extractive one: what happens to the numbers when the thing that made them goes away.

What this case doesn't prove

Some of the recovery is timing rather than skill — live, weekly, year-round programming became structurally scarce at exactly the moment streaming platforms started competing for it, and that would have raised the price of these rights regardless. Some of the record gate figures reflect higher ticket prices rather than higher demand; one analysis found average prices for the company's regular North American shows rising from roughly $75 in 2024 to roughly $118 in 2025. And there are signs of strain in the weekly touring business since, which is what the extraction pattern looks like when it begins. This is one company, in an unusual category, over one five-year window. Treat it as a mechanism worth testing, not a law.

The mechanism is this. A revenue floor is a tool, and the tool has two settings. It can fund the patience that lets a brand compound, or it can fund the years in which nobody has to look at what's breaking. It does not tell you which setting it's on. You have to build the instruments that do — and you have to build them while the numbers still look good, because that is the only window in which anyone will let you.

Most organizations never get the correction. Not because they lack conviction, but because their floor is high enough to postpone the reckoning and not high enough to survive it.

Related from Sound Decisions: Value Creation Is a Business Model, Not a Moral Posture · Additions Are Free — Which Is Exactly Why They Don't Work · The Spend You Inherited Is Two Different Problems · The Ground Nobody Was Defending

AI Doesn't Give You Leverage. It Multiplies the Process You Already Have.

Nearly every agency now uses AI, and the most common operational complaint is still inefficient process. The two facts explain each other: the tools multiply the process a firm already has, and most firms have less of one than they think.

If you are tired of being told you are using AI wrong, that is a reasonable place to stand. The seats are bought. The team drafts with it every day. You have watched a demo turn a week of work into a morning, and watched your margin stay exactly where it was.

Between those two facts sits a suspicion that the missing piece is a better prompt, model or vendor. The evidence points somewhere else. AI does not create leverage on its own. It multiplies the process a firm already has, and most firms have less process than they think.

The original version of this piece was built on a long interview with the owner of a done-for-you agency running about fifteen accounts in a single construction-adjacent vertical. His stack works. Onboarding produces a complete brand and campaign foundation from a fixed sequence of prompts run against an intake form, the client's site and their customer list. Building a campaign went from about a week to half an hour.

Then the part that matters. Before any of it, he spent six and a half years on the phone, averaging around 250 dials a day. He can describe the contact sequence he later automated hour by hour, because he ran it himself. The machine did not discover his process. It inherited one. As he put it, the AI did not make him good at follow-up; it let him stop personally being the follow-up. Only one of those is available to a firm that never wrote its process down.

The obvious objection is that well-run firms adopt technology well, so process discipline may be a symptom of competence rather than a cause. One clean test exists. In a randomized field experiment published in the Quarterly Journal of Economics in 2013, Nicholas Bloom and colleagues gave free management consulting to a randomly chosen group of large Indian textile plants and compared them with controls. Adopting structured practices raised productivity about 17 percent in the first year.

The detail worth keeping: computer use went up afterward, because the new processes demanded data that paper could not supply. The process came first and pulled the tools in behind it. Asked why the plants had not adopted profitable practices on their own, the researchers pointed mainly to information. Owners did not know what good looked like.

McKinsey's survey data points the same way at the scale of large companies. In its March 2025 State of AI report, fundamental workflow redesign had the biggest effect on whether organizations saw bottom-line impact from generative AI, out of 25 attributes tested. Only 21 percent of adopters had done it. By its November 2025 edition, just 39 percent of respondents attributed any enterprise-level profit impact to AI, and the roughly 6 percent it classed as high performers were far more likely to have redesigned workflows.

Agencies are not underusing AI. They are using it precisely where using it changes nothing.

One interview is an illustration, not proof. A successful operator's account of why he succeeded is filtered through survival and assembled after the fact. The owner and firm are kept anonymous by agreement.

The randomized evidence comes from textile plants, not agencies. It supports the order of operations; how far it carries into creative services is a judgment.

McKinsey's model is correlational. Workflow redesign was the strongest of 25 attributes, and together they explained about a fifth of the variation. It is the best single lever on the list, not a guarantee. Basis sells automation software to the agencies it surveys.

Writing a bad process down buys a machine that runs a bad process faithfully. Documentation is necessary, not sufficient. The Bloom result is about good practice, not any practice.

Being a good strategist is not hard because it is clever. It is hard because it has to be repeated, for the next account and again next quarter for all of them at once, by a tired person. That is exactly where a written process and a patient machine earn their keep, and exactly where most firms have neither.

Related from Sound Decisions: Creative Is Now a Line Item · You Don’t Need More Clients. You Need Longer Ones. · They’ll Pay for the Work. They Won’t Pay for the Thinking.

Anyone Can Look Like You Now. What is left is what you know.

Logos, awards, case studies and credentials can now be produced without the work underneath them. When proof gets cheap, buyers discount all of it, and the discount lands hardest on the firms that earned theirs.

Start by conceding the whole thing. Your logos are real. You won those accounts, the work shipped, the results happened, and the people who did it still work for you. The instinct that the fakes are over there and you are over here is reasonable.

It is also exactly how the problem reaches you. A signal carries information only while it is expensive to produce. Once it can be produced without the thing underneath, it stops telling anyone apart, and it stops telling you apart first, because you are the one who paid full price for it. This is not a story about bad actors. It is a story about a cost collapsing, and what that does to the firm that paid.

It happened in several places at once, for different reasons, and from where a buyer sits it looks like one thing.

In company filings, the adjusted number became the house style. Calcbench and Suffolk University's July 2026 study found 361 companies in the S&P 500, 72 percent of the index, reported adjusted earnings for fiscal 2025, and their adjusted net income ran $271 billion above net income under standard accounting rules. Eighty-seven percent of the adjustments moved earnings up. Little of this is improper. It does mean a reader can no longer take a headline number at face value.

In reviews and testimonials, the floor fell out. The Federal Trade Commission's rule on consumer reviews, in force since October 21, 2024, bans fake and AI-generated reviews and lets courts impose civil penalties for knowing violations, now up to $53,088 each. The FTC's own description of the harm is the mechanism in one line: fake reviews pollute the marketplace for everyone, including the businesses that never bought one.

And on agency websites, the logo wall has always been loose. Past clients, clients won by people before they joined, marks from a single project years ago: all of it sits on the same wall with the same weight. None of the three shifts is a scandal. Each is a norm drifting, one page at a time.

Sort every claim on your website into two piles, using one question: could someone who has never done the work produce this?

When a class of signal becomes unreliable, buyers keep buying. They discount the whole class and look for something else to trust. The discount does not check who earned the credential, which is why it costs the honest firm most: real logos and borrowed ones now reach the same reader carrying the same, reduced weight.

What clients reach for instead shows up in why they leave. In Setup's 2025 survey, 44 percent of clients who ended an agency relationship said the agency did not understand their business. Asked the same question, 3 percent of agencies named it. In 2024, delivery problems had been the top reason, cited by 48 percent. Clients are grading on what the agency knows and does. Agencies are still presenting what they have done.

The encouraging part: independent agencies already start from the strongest position on the second pile. The ANA and the 4A's reported in April 2025 that independents keep agency-of-record clients 7.3 years on average, against 5.8 for the holding companies. That lead is built out of knowing clients, not out of the wall.

None of the three trends is about agencies directly. Adjusted earnings and fake reviews are evidence that proof in general is getting cheaper, and that regulators and standard-setters are reacting. That buyers discount agency credentials as a result is an inference, supported by the churn data but not measured on its own.

Setup's figures are self-reported and sold alongside a search service, and the ANA and 4A's tenure figures are self-reported too.

Knowing the business does not rescue bad work. Understanding is a tiebreaker between competent firms, not a substitute for competence. If the work is late or wrong, insight will not hold the account, and it should not.

It does not scale by effort. Deep knowledge of a client's business is sustainable across one or two accounts. Across five it becomes a second job, carried quietly until the person carrying it runs out.

A credential is worth what it costs to obtain. When that cost falls, the credential falls with it, including the ones that were earned. What cannot be produced without the relationship is the only proof whose price still holds.

Related from Sound Decisions: Your Best Clients Are the Ones You Undercharged · Nobody Pays You Extra for Being Honest · They’ll Pay for the Work. They Won’t Pay for the Thinking.

When the Client Wins, They Win. Three tests for an agency whose fate is bound to its client's.

Most agency rankings count awards or revenue. Neither tells you whether an agency wins only when its client does. Three tests do: a client-credited result, a relationship that survived a real test, and a fee that moves with the client's outcome. The third is almost never public.

In 2018, Shiner Beer put its creative account into review. It had been with the Austin agency McGarrah Jessee for about fifteen years, and on the way out the brewer's marketing director said on the record that the agency had overseen the period with the largest sustained growth in Shiner's history, and that the two were parting on very good terms.

If you run an agency, sit with that for a moment. The work had done what work is supposed to do. The client said so, publicly. The account left anyway.

Most owners believe good work is what protects an account, and every agency ranking reinforces the belief, because rankings count awards, revenue or noise. None of that says whether an agency and its client actually win together: whether, when the client's business grows, the agency helped grow it, and when the client struggles, the agency is in it with them.

That reciprocity is what separates a partner from a supplier, and it can be tested. Three questions do most of the work. The public record offers clear Texas examples of the first two, and the Shiner story, which did not end in 2018, is one of them. The third is almost never visible, which is itself a finding.

Not an award; a result. Shiner's exit statement is the example. A client crediting growth to an agency on the way out is rarer than one doing it on the way in, and it is the one kind of evidence no agency can write for itself.

A new chief marketer, a formal review, a recession, an acquisition. Vendors get cut in those moments; partners get kept.

Houston's Lopez Negrete Communications began working with Walmart in 1995. When Walmart ran a broad agency review starting in 2005, it replaced all of its general-market agencies. Lopez Negrete kept the business, and Ad Age reported in 2012 that about 80 of the agency's 200 staff worked on it.

And Shiner is the other side of the same test. In January 2026, after seven years with other agencies, the brewer brought its creative account back to McGarrah Jessee. Its brand director said it wanted a partner who understood the brand, and credited the agency's commitment and perseverance.

Results-linked fees, a public commitment, walking away from safe billing to protect a client's result. This is the strongest signal of all, and the original index could not find a single Texas independent with public evidence of it. That does not mean it does not happen. It means agencies and clients do not publish it, and so the market cannot see the one thing that would most clearly separate a partner from a supplier.

Longevity alone is not the point. Many relationships last because nobody has tested them. What the two cases share is that each was tested, publicly, and held, or broke and was mended. That is the difference between a client who stays and a client who chooses you again.

The best evidence is usually private. A revenue lift or a retention save is known to the agency and the client and rarely published in a form anyone can check. Public examples are the visible fraction, and many strong agencies are absent from the record for that reason alone.

Public statements are selected. Clients praise agencies in announcements and exit statements; they rarely publish criticism. A credited result is strong evidence, not proof.

Relationships change. McGarrah Jessee's creative relationship with Whataburger went into review in September 2025; its outcome was not public in the sources read for this piece. Any example can be overtaken by events, and the examples here are dated.

This is not a ranking. The original index sorted agencies into tiers. That has been retired; a list nobody maintains goes stale and turns into a claim no one is checking.

The question was never who does the best work. It is who wins only when the client wins, and whether anyone outside the relationship can see it.

Related from Sound Decisions: You Don’t Need More Clients. You Need Longer Ones. · You Can Refuse the RFP and Still Win Clients · Your Best Clients Are the Ones You Undercharged

Your Best Clients Are the Ones You Undercharged.

Every client relationship runs a quiet balance: the client gets more than they pay for, or less. The accounts tilted toward the client are the ones that last, and the agencies losing the others are usually looking for the reason in the wrong place.

“I can’t give work away. That’s how an agency goes broke.” Most owners hold some version of that, and they have reason to. Margins are thin, every hour has a cost, and generosity looks like the fast road to a business that works hard and keeps nothing.

Grant it completely. An agency that leaves value on the table with the wrong client, at the wrong price, on the wrong terms, does go broke. What follows disputes something narrower: the belief that the safest thing to do with a good client is to capture as much of the relationship as the contract allows. On the evidence, that instinct is the expensive one.

Strip a client relationship down and there is a ledger under it. On one side, what the client pays. On the other, what they actually get: the outcome, the judgment, the hours nobody billed, the problem caught before it became theirs. When the second side is heavier, the client carries a surplus. When the first side is heavier, the agency is extracting.

That balance moves every quarter, and the client feels which way it is tilted long before either side says anything. Surplus is not the same as a discount. You can hold your price and still tilt the balance toward the client by giving them more of the one thing they cannot get elsewhere: your judgment about their business.

Value has been at or near the top for a while. In Setup's 2023 survey, dissatisfaction with value became the leading reason clients ended a relationship, rising from 39 percent to 53 percent in a year. And the agencies have been misreading it for as long as Setup has asked. In its 2019–2020 survey, 75 percent of agencies believed budget cuts were the main reason relationships ended. Only 24 percent of clients said so.

Almost nobody decides to start extracting. It happens by drift. Margin tightens on an account, so the agency protects it by quietly doing a little less rather than pricing the value higher. A strong piece of thinking gets billed as a line item, or given away and never mentioned again. Contact narrows until it clusters around scope and renewal. Each step feels like discipline. Together they tilt the balance, and the client notices before the agency does.

That is why a review so often arrives as a surprise. Nothing looked wrong, because nothing was said to be wrong. Setup puts the pattern plainly in its own summary: relationships deteriorate as missed expectations accumulate and value becomes harder to explain, rather than over one bad meeting.

The retention argument usually leans on two famous figures: that winning a new customer costs five to 25 times more than keeping one, and that a 5 percent gain in retention lifts profit by 25 to 95 percent. Both trace to Frederick Reichheld's work at Bain, which began with a 1990 Harvard Business Review study of banks, insurers and an auto-service chain. When HBR repeated the cost range in 2014, it said the figure depends on which study you believe and what industry you are in. Treat them as direction, not measurement.

The agency-specific evidence points the same way and is more recent. Promethean Research's 2026 growth guide calls account management the highest-leverage revenue function it measures, and reports that clients who arrive by referral stay about 1.9 times longer than clients won through events, networking or outbound. It also finds that a quarter of agency engagements end within a year. In April 2025 the ANA and the 4A's reported that independent agencies keep agency-of-record clients 7.3 years on average against 5.8 for the holding companies, and that relationships without mandatory reviews last 8.1 years.

Headcount buys revenue. Tenure buys profit. And tenure is bought with a balance the client can feel, not with a tighter scope.

Setup runs agency searches and publishes its survey partly to market that service. Its sample size is not in the summary, and several reasons can be cited at once, which is why value and delivery can both reach 61 percent. Read the ranking and the gap between clients and agencies; weigh the exact percentages lightly.

The retention multipliers are old and general. The Bain figures come from consumer financial services in the 1990s. They support the direction of this argument, not its size.

Surplus does not rescue a relationship that is underwater. If an account is mispriced or behind on the work, leaving more value on the table loses money faster. That is a pricing or delivery problem. This argument is for the relationship that is sound and quietly drifting, which is the common and expensive case.

Surplus poured into one large client builds dependency. A book of balanced accounts compounds. One generously served account just concentrates the risk.

The clients agencies keep longest are rarely the ones they charged most. They are the ones who could always say what they were getting. The balance you cannot see on any invoice is the one that decides the renewal.

Related from Sound Decisions: They’ll Pay for the Work. They Won’t Pay for the Thinking. · Value Creation Is a Business Model, Not a Moral Posture · Your Next Client Is Already Inside This One

Value creation is a business model, not a moral posture — and it beats extraction on the numbers

Every company makes one quiet decision before it makes any loud ones: how much of the value it creates it intends to keep. The operators who chose to keep less, on purpose, built the harder businesses to kill.

Every pricing meeting has the moment. Someone looks at a margin, or a fee you have never charged, or a discount you have always given, and says: we're leaving money on the table. Heads nod. Disagreeing would mean arguing for taking less, and nobody wants to be the one in the room who sounds soft.

What makes the moment so easy to lose is the belief underneath it, usually unspoken because it is treated as too obvious to state: the way you make money is to take more than you give. Take it from the customer through the markup, from the supplier through the squeeze, from the people who do the work through the terms. Grow up, the belief says. Business is extraction, and anyone who tells you otherwise is either naive or losing.

I think that belief is the naive one, and I think it is expensive. Not because generosity is virtuous, but because the arithmetic runs the other way over any real length of time. What follows is three companies that faced the same decision — how much of the value they created to keep — and made it differently. The decision came years before the outcome. The outcome is the part everyone remembers; the decision is the part an operator can actually copy.

A ceiling, set on purpose

The cleanest example of the value-creation decision is not a mission statement. It is a cap.

The trap most retailers fall into is obvious once you name it: the retailer profits by charging more, so the retailer's interest is set against the customer's. Every shopper feels that, which is why they comparison-shop and why loyalty is thin. Costco removed the conflict. It decided not to make its money on the markup at all — nearly all of its profit comes from the annual membership fee. The low prices stopped being a cost to manage and became the thing that makes the membership worth renewing.

Now sit inside a competitor's numbers. To beat Costco on price, a rival has to profit from the exact markup Costco gave up. The generosity is not softness; it is a wall the extraction model cannot climb. This is the whole argument in one company: refusing to take the last dollar is not a concession. It is the moat.

The decision, in three rooms

"No, there are no in-game purchases in our game."

Larian Studios, Baldur's Gate 3 launch FAQ, 2023

The mechanism, not the sermon

This is a structural argument, not a moral one. Value creation beats extraction for reasons you can put in a model.

Taking less is not the creation. Start with the value a business generates: the gap between the most a customer would pay and the least its suppliers and staff would accept. Price decides how that gap gets divided — and moving price only moves value between the parties. It makes none. So what Costco did was subtler than restraint. Capping the markup made the membership worth renewing, which raised what a member would pay for the one thing Costco does charge for. Larian’s refusal to sell the game twice raised what a player would pay for it once. The restraint is the visible part. The creation is what the restraint bought — a higher ceiling on what people will pay, and a lower floor on what it costs to keep them.

Goodwill is asymmetric. You can buy it — it sits on the balance sheet as the premium an acquirer pays over the hard assets. But that is exactly the tell: purchased goodwill is what gets written down when the new owner starts extracting, because the number turns out to be worth less than they paid. The goodwill that holds is earned, one kept promise at a time. A competitor can copy your price, product, and features overnight; they cannot copy fifteen years of a customer trusting you'll pass the savings on. You can spend that trust in a quarter. You cannot rebuild it in one.

Surplus replaces paid acquisition. When customers get more than they paid for, they tell people — the cheapest, highest-trust growth there is. Trader Joe's runs on it: around 80% private label, no real advertising, no loyalty card, a growth engine that is entirely customers who love it enough to talk. And the platform proves it at scale — YouTube reports paying creators, artists, and media companies more than $100 billion over four years, handing the majority of ad revenue to the people who make the videos. It removed the toll between a creator and an audience, and built one of the most valuable media businesses of its generation doing it.

There is an honest limit worth stating, because a decision aid that overclaims is useless. This model is easiest to hold when the owner is patient or the business is protected from short-term financial pressure. Sinegal guarded the cap. Larian answered to no outside owner. The BuzzFeed creators had to leave to build on their own terms. The real fight is not in the market — it is on the ownership structure, and it starts the day someone with a shorter time horizon can overrule the cap. Build the business to create value, then build the ownership to defend it.

Extraction optimizes the next transaction. Value creation optimizes the relationship the transactions come from. Over any real time horizon, the relationship wins — and that is a decision you make once, at the start, and then have to defend from the version of yourself who someday gets tired of being generous.

Related from Sound Decisions: Your Best Year on Paper Can Be the Year the Brand Starts Dying · The Marketing Job Is Derived Power · Additions Are Free

Your Velocity Number Means Three Different Things

Units per store per week. Sales per point of distribution. Dollars per million ACV. All three are called velocity, all three are in daily use, and they can point in opposite directions for the same brand in the same quarter. Your board hears one number.

The quarter closed up twenty-two percent and the room is pleased. Someone puts the number on a slide with an arrow on it. Then a buyer at your second-largest account tells your sales lead the brand is underperforming the category and they are looking hard at the set in the next line review.

Both statements are true. They are describing the same twelve weeks. And whichever one you repeat in your next board meeting will shape what the company does for the following two quarters.

This is not a data-quality problem, and it will not be fixed by better dashboards. It is a definition problem, and it is old, and almost nobody in a CPG company will name it out loud because everyone assumes everyone else means what they mean.

Three measures, three questions

Velocity, in the broadest sense, means the rate a product sells where it is available. The disagreement is about the denominator, and the denominator is the whole argument.

Units per store per week divides units sold by stores selling and by weeks. It gives you a figure a buyer can hold without a calculator — this item moves about six units a week in an average store — and it is the natural unit for replenishment and for the question every line review actually asks, which is whether an item earns its facing. Its weakness is that it treats every store as equivalent. A unit sold in a large supercenter counts the same as a unit sold in a small natural grocer.

Sales per point of distribution divides sales by the percentage of distribution achieved. It is built to answer a different question: is the brand getting more productive as it expands, or just bigger?

Dollars per million ACV normalizes by the store's all-commodity volume — its total sales — so that performance is weighted by how much traffic a location actually has. It exists precisely to remove the distortion that units per store per week leaves in.

Each is correct. Each answers a real question. None of them is the velocity number, and any deck that presents a single figure labeled "velocity" has quietly made a choice on your behalf.

The arithmetic is not subtle once you see it. Add stores faster than each store's rate of sale holds up, and the total climbs while the per-store figure sinks. Total sales is the number that makes a growth story. Per-store rate of sale is the number a buyer uses to decide whether you keep the space.

Which means a brand can post its best quarter and lose a listing in the same month, and everyone involved can be reading their own report correctly.

The retailer is not using your number

Worth being blunt about this, because it is the part that costs money.

The buyer is not evaluating your growth. They are evaluating the productivity of a piece of shelf, and the unit that governs that decision is closer to margin dollars per shelf foot per week than to anything on your dashboard. Every linear foot has an expected return. If your item is not clearing the threshold for its space, it gets replaced by one that does, and your total-sales growth is not an argument the buyer is obliged to care about.

So rising distribution with softening rate of sale is not a mixed result. It is a leading indicator of delistings, arriving one or two line reviews ahead of the delistings themselves. It is the single most reliable early warning available to a CPG marketing leader, and it is invisible to anyone reading total sales.

Row five is not a criticism of boards. It is a description of what happens when nobody supplies anything better.

And nobody owns it

Here is the organizational half, and it follows from the measurement half rather than standing on its own.

Sales owns the retailer relationship, which means they own the conversation where velocity decides something. Marketing owns awareness, trial, and the demand that produces velocity in the first place. Finance owns the P&L, where velocity shows up only after it has already turned into revenue or its absence. Category management, if the company has it, owns the syndicated data and usually reports into sales.

So the number that determines whether you keep your shelf, what your next round is worth, and how long your tenure runs sits in the space between four functions, and belongs to none of them. Each team reports the version that flatters its own work, and each is being honest, because each is using a legitimate definition.

An unowned number is an unmanaged number. It will still be attributed to you.

What to do about it in your first quarter

Two moves, and the first is nearly free.

Pick the definition and write it down. One measure becomes the company's velocity number — units per store per week is the usual right answer for a growth-stage brand, because it is what the buyer is using and the buyer holds the decision that matters most. The others stay available and stay labeled. Anyone presenting velocity says which one they mean. That is a fifteen-minute decision that removes a recurring, expensive ambiguity.

Then claim it or name its owner, in writing. Not a dashboard, not a working group — one person accountable for the trend, reporting it the same way every month, in the same units, with distribution shown alongside so the two can never again be read apart. If that person should be you, say so early, while it still looks like leadership rather than like a land grab. If it should be sales, say that too, and make sure the reporting reaches you unfiltered.

The claim under both moves is the one worth carrying out of this piece. You are accountable for a number you do not control, defined three ways, computed differently by the party who decides your shelf. You cannot fix all of that in a quarter. You can make it visible, which is the part that has to happen before anything else can.

Most of Your Marketing Budget Is Rent

A large share of what you are accountable for is not marketing at all. It is contra-revenue — money that never reaches the top line, paid to retailers for access. Until you separate the rent from the demand, every efficiency conversation you have is about the wrong number.

Somebody slides a number across the table and calls it the marketing budget. It is large, it is yours now, and the expectation attached to it is that it will produce growth in proportion to its size.

Open it up and a substantial share of that number is not buying anything a consumer will ever see. It is slotting fees. Off-invoice discounts. Temporary price reductions. Retailer deductions and chargebacks you did not authorize and cannot easily contest. It is the price of being on a shelf and staying there, negotiated between your sales organization and a buyer, and charged to a line with your name on it.

That is not a complaint about the sales team. It is a category error, and it is sitting in the middle of the number you will be judged on.

The accounting already disagrees with the org chart

Here is the part most marketing leaders are never walked through, and it is not a matter of interpretation. Trade spend is treated as contra-revenue. It does not sit in operating expenses beside your media budget. It reduces gross sales on the way down to net sales, which means it disappears from the top line before anyone gets to the section of the P&L where marketing lives.

The practical consequences are worth stating plainly, because each one lands on you.

Booking trade spend as a marketing expense rather than as contra-revenue overstates both revenue and gross margin at the same time. Every decision made downstream of those two figures — pricing, hiring, production runs, what you tell an investor about unit economics — gets made on numbers that are not real. This is common enough in growth-stage brands that it is worth checking personally rather than assuming.

It also means a large share of your accountable budget never appears as marketing spend anywhere in the financial statements. It appears as revenue the company never collected. You are answering for money that, in accounting terms, was a discount.

Read that source line again

You will hear that figure quoted at you. It will be attributed to McKinsey, confidently, in a pitch meeting, probably by someone selling trade promotion software.

It is not McKinsey's finding. McKinsey published it in 2019 and footnoted it to a Nielsen report from 2016. Trace the citations on a dozen vendor pages and they converge on that single 2019 article, which is itself a citation. The number a marketing leader will be handed as current consultancy research is a decade old and belongs to somebody else.

It is also narrower than the way it gets used. A promotion that lost money is not the same as a promotion that did nothing — some bought trial, some defended a facing, some were the price of a listing that carried the rest of the year. And it is a promotion-level finding being deployed as a judgment on entire budgets.

Worth sitting with, because it is the same problem one level up. If the headline number about the second-largest line item in your industry can travel for ten years without anyone checking its parentage or its scope, be careful what you accept about your own lines from people with something to sell.

Splitting the number

The work is not complicated. It is just nobody's job, which is why it does not get done.

Rows four and six are where the argument happens, and they are the rows worth your attention. Retail media in particular is funded out of trade budgets at most companies while behaving like advertising, which means it is measured by whichever team happens to claim it. That ambiguity is convenient for everyone and useful to no one.

Do not read the map as a hierarchy. Rent is not waste. A brand with no shelf has no business, and access has a price like anything else. The point is that rent and demand answer to different logic. Rent is negotiated, and you improve it by negotiating better, auditing deductions, and walking away from doors that cost more than they return. Demand is built, and you improve it by making something people come back for. Running both through one number, judged by one metric, guarantees you manage at least one of them badly.

Why this lands on you and not on finance

Finance already knows. The contra-revenue treatment is in the filings; the gross-to-net walk is a standard slide. Nobody is hiding anything.

What is missing is that the person accountable for the blended number is usually the person with the least authority over its largest component. Trade terms are negotiated by sales. Deductions arrive from retailers. Your leverage sits almost entirely in the demand half — and the demand half is the smaller half.

So when the board asks why marketing efficiency is not improving, and the number they are looking at is mostly rent, you will find yourself defending a figure you do not control with tools that only reach part of it. That conversation does not get better with effort. It gets better with a different number.

Which is the whole ask here, and it is a reporting change rather than a strategy: split the budget in two and report them separately from your first board meeting onward. Rent, with its own owner and its own trend line. Demand, with yours. Then the efficiency question has somewhere to land, and the answer to "why is this not working" stops being a story and starts being a location.

The Spend You Inherited Is Two Different Problems

Every inherited marketing budget holds two kinds of spend — the kind that cannot prove it works, and the kind you cannot see yet. They look identical on a spreadsheet. Cutting them the same way is the most expensive mistake available in your first quarter.

The money is in the bank. The plan attached to it assumes most of it leaves this year — into national doors, into the trade calendar, into a media budget larger than anything this company has run before. And the spending history you are meant to build on top of is a spreadsheet you did not write, describing decisions you were not in the room for.

Nobody left a note explaining why the creator program exists, or which of the four trade shows was the one that mattered, or whether the agency retainer is buying strategy or invoices. Some of the people who could tell you have moved on. The ones still here have reasons to describe the past a particular way. You have a number to hit, a team watching to see what you cut first, and a strong instinct that a good portion of this is waste.

That instinct is probably right. It is also the most dangerous thing you are carrying into your first budget cycle.

Two piles that look identical on a spreadsheet

The discipline that governs this decision is simple to state. Every line of spend is either feeding velocity or feeding the appearance of marketing, and one question settles it: did it move units per store, per week? If you cannot tell, it is a candidate for the cut.

That rule is unforgiving on purpose, and for a founder who built the spend themselves it works cleanly. They know what they bought and why. When they cannot tell whether something worked, the not-knowing is usually the answer.

You are in a different position. In your first ninety days you cannot tell about almost anything, and the reason has nothing to do with whether the spend works. It has to do with when you arrived.

So the inherited budget splits into two piles that look exactly the same in a spreadsheet and are not the same at all. There is spend that cannot prove itself — the measurement exists, or could exist, and it shows nothing. That is a verdict. And there is spend you cannot see yet — it may be working, but the reporting was built by someone who needed a particular story to be true, or was never built at all. That is not a verdict. That is a task.

Cut the first pile and you free cash. Cut the second pile and you are guessing, in public, with the board watching, about lines you never understood. Some of what you kill will have been the only thing working, and you will not find out until it is too late to reverse.

That gap is the whole problem. Your clock is shorter than the feedback loop on the decision you are about to make. Nothing about the incentives will help you here, which is why the discipline has to.

When you can finally see it, a lot of it really is nothing

The case for cutting hard is strong, and the best evidence for it comes from the largest advertiser in the world.

Read that carefully, because the sequence matters more than the headline. P&G did not cut and then look. It built the ability to see first, and the cut followed as an obvious consequence. The transparency campaign came before the reduction, and the second reduction came only after the first one proved harmless.

"We were dazzled by the shiny objects and big data overwhelmed us."

Marc Pritchard, P&G, on the years before the cut · 2018

And sometimes you cut because you could not see, not because it was not working

The other case is two capable companies reading identical uncertainty in the same weeks and reaching opposite conclusions.

Be careful what you take from this. Coca-Cola's 2020 revenue collapse was driven by restaurants and stadiums closing, not by its ad budget, and anyone who tells you the cut caused the decline is selling something. That is not the point. The point is the reasoning. "I cannot see a return right now, so I will stop" treats an absence of measurement as a finding. It is the most natural sentence in the world to say in a hard quarter, and it is the exact error this piece is about.

Notice too that P&G appears on both sides. In 2017 it cut hard, because it had built the ability to see and what it saw was nothing. In 2020 it spent into a fog, because it could not see and refused to treat that as evidence. Same company, opposite actions, one principle underneath: act on what the measurement says, never on whether the measurement exists.

You were hired to deploy, not to prune

There is a second pressure working against you, and it comes from the round itself.

A Series B is raised on a growth story, and the plan attached to it assumes the money goes out the door. Walking into your first board meeting with a list of things you intend to stop can read as timidity, or worse, as a marketing leader who does not believe in marketing. That reading is available to anyone who wants it, and someone in the room usually does.

The way through is not to soften the cut. It is to state what the freed money is for. A budget that concentrates reads as conviction. A budget that spreads reads as hedging — as a leader who could not decide, so funded everything a little. Boards are considerably better at recognizing the second thing than marketers assume.

And the alternative is worse than it looks. The fastest available route to becoming the third departure is to inherit the second one's budget, add the new round on top, and spend eighteen months unable to say which part of the total did anything.

The question that sorts the piles

There is one question that separates a verdict from a task, and it is not "is this working."

It is this: can you name what would have to be true for this line to be working, and can you get that number within sixty days?

If you can name it and get it, the line is not unprovable. It is unseen. Put a number and a date on it, tell the team both, and let it run to the deadline. You have converted a guess into a scheduled decision, which is the only real trade a new leader can make against a short clock.

If you cannot name what proof would even look like — if nobody in the building can describe the outcome this spend is supposed to produce, in units, in a store, in a week — then the line is not merely unmeasured. It is unmeasurable as currently constructed, and it was never going to earn its place. That one you can cut in your first quarter and defend in any room.

Most inherited budgets, run through that question honestly, come apart into a small number of lines somebody can defend with a number, a larger number of lines somebody defends with a story, and a handful nobody defends at all. Start at the end of that list and work backward. The order protects you, and it also happens to be the order that produces the most cash for the least risk.

You Can Refuse the RFP and Still Win Clients. The pitch is where you accept the terms.

The RFP is not where an agency wins the work. It is where it accepts the terms, at the moment it has the least leverage it will ever have with that client. Two questions on the first call change the economics more than the fee does.

Nothing here argues for wanting smaller clients. The email from a national brand, with a budget that would change your year, is not a fantasy, and this is a better year for independents to get it than most. The question is what you agree to on the way in, because the process that hands you the logo also hands you the terms, and almost nobody reads the second part.

The cost of entering a pitch is covered in an earlier piece in this series: about $204,461 for a non-incumbent agency, with the incumbent kept two times in three. This one is about what happens if you win.

Most of an agency's costs are people, and payroll comes due every month. So do rent and, increasingly, freelancers: New York State's Freelance Isn't Free law, in force since August 2024, requires written contracts and payment within 30 days for covered freelance work. When a client pays in 60, 90 or 120 days, the gap does not disappear. The agency finances it. The client's cash position improves by exactly what the agency's worsens.

The agency trade body has said so without hedging. In its 2023 guidance on extended payment terms, the 4A's stated that anything beyond a 30-day cycle is incompatible with the typical agency commercial model, and called fronting media costs for clients particularly destructive, turning agencies into banks for clients that are usually better capitalized.

The argument here is about leverage rather than a measured rate. An agency has the least leverage it will ever have with a client at the moment it is trying to be chosen. Forrester has estimated the agency review process costs the industry about $12.5 billion a year, and reported that 76 percent of chief procurement officers prioritize reducing cost. A term sheet presented inside a review, run by people whose stated priority is cost, is accepted under different conditions than the same terms raised at a renewal, after two years of results.

Big clients do not leave faster. Reviewed clients do. The ANA and 4A's found average agency-of-record tenure is now about seven years, independents keep clients longer than the holding companies, and the variable that separates long relationships from short ones is whether the client re-tenders on a schedule. The review is not a one-time cost. It installs a clock.

The payment-terms data is old. The ANA's figures are from 2019. The 4A's cited later studies putting average terms near 60 days; nobody has published a recent U.S. figure at the same standard, and practice may have moved further.

That terms are set worse inside pitches is reasoning, not measurement. The leverage argument is strong and consistent with the tenure data, but no study has compared terms accepted in reviews with terms negotiated at renewal.

Some pitches are worth it. A review with no incumbent, clear terms and paid participation is a different decision from a defended incumbency with free creative requested. This piece argues for asking before entering, not for never entering.

Freelancer payment law varies by state and city. New York's is cited because it is statewide and specific; check the rules where your freelancers work.

You do not need to win more pitches. You need to be in fewer of them, to know the terms before you walk in, and to be the obvious answer before a review is ever called.

Related from Sound Decisions: Your Next Client Is Already Inside This One · They’ll Pay for the Work. They Won’t Pay for the Thinking. · The Budget Line

Poppi Didn't Go Viral by Accident. Two Decisions Came First.

Everyone saw the pink cans and the Super Bowl ads. The exit was decided years before any of it — in two choices most founders never get to make on purpose.

You have probably had this thought. You're scrolling, and there's Poppi again — a hot-pink can in a celebrity's hand, a Super Bowl spot, a video with a few million views. The brand sold to PepsiCo for close to two billion dollars. And somewhere underneath the admiration is a quieter, less comfortable question: is that the answer? Do I need to go viral?

Right behind that question is a fear most founders don't say out loud — that if the answer is "go viral," then the answer is really "get lucky," and luck is not a plan you can run.

Here's the good news. The virality wasn't the plan. It was the result of one.

The lesson everyone takes from Poppi is the wrong one

The story people tell about Poppi is a story about culture: the influencers, the pink cans, the Super Bowl. It's a good story. It's also the wrong lesson, because it describes what you could see, not what caused it.

Poppi didn't become worth buying because it went viral. It went viral because, years earlier, two decisions were made that most brands never make on purpose. The culture was downstream of those decisions. And unlike luck, decisions are something you can study — and sometimes copy.

One thing the popular version leaves out: this wasn't only a founder's kitchen tale. Allison and Stephen Ellsworth started the drink, and that origin is real. But the two decisions below were made with an investor-operator in the room — Rohan Oza and his firm, CAVU — who had built beverage brands before. That matters, because it tells you these were craft, not accident.

1 · Sell an occasion, not an ingredient

Poppi didn't start as Poppi. It started as Mother Beverage — an apple-cider-vinegar drink, sold in glass bottles, positioned as a health tonic, living where health tonics live: the natural-foods aisle.

That aisle is a door. For a lot of better-for-you brands, it's the wrong door — a narrow, slow, self-limiting shelf where you're bought by people who already care about apple cider vinegar and ignored by everyone else. A drink defined by its ingredient can only ever be as big as the number of people already shopping for that ingredient.

Around 2020, with Oza and CAVU, the brand made its first real decision: stop selling an ingredient, start owning an occasion. Mother Beverage became Poppi. The glass bottles became bright cans. The apple-cider tonic became a soda — something you reach for in the moment you'd otherwise reach for a Coke, minus the thing you were trying to avoid.

That's not a marketing change. That's a change in what the product is for. A tonic is a thing you take. A soda is a thing you crave. One ages out with a trend; the other owns a moment people come back to every day. Poppi stopped competing for the small shelf of people who wanted vinegar and started competing for the enormous, durable occasion of "I want a soda I don't have to feel bad about."

One honest caution, because this is exactly where founders over-learn the lesson: creating a category is not a formula. Most attempts to create a new category fail — quietly, expensively, and without a Super Bowl ad to show for it. Poppi is one winner, and survivors always look inevitable in hindsight. Read this as a path that exists, not a path that works.

2 · Build the demand before you fight for the shelf

The second decision is the one the culture story actually points at — but it usually gets the order backwards.

Most brands chase distribution first. They fight to get onto the big national shelf, and then they hope demand shows up to justify the space. Usually it doesn't, and the brand quietly gets removed — because a shelf you can't move product off of is a shelf you lose.

Poppi ran it the other way. It built the demand first, before the shelf, so that by the time it earned major national distribution the pull already existed. The video that reportedly did a hundred thousand dollars in sales in a single day. Back-to-back Super Bowl ads. Celebrities holding the can. That machine wasn't vanity. It was manufacturing the one thing that keeps you on a shelf: people walking in already wanting you.

So when Poppi landed a deal across roughly 4,600 Walmart stores, it did something telling: it didn't fight for space next to Coke and Pepsi. It carved out a new set — a functional-soda space, one it shared with rivals like Olipop, that hadn't existed at that scale before. That only works if the demand is already there. You don't get to define your own shelf until enough people are asking for you by name.

That's the sequence, and it runs in exactly one direction: build the demand, and people arrive already wanting you; the shelf holds because it moves; and a shelf that holds is what lets you scale nationally on your own terms — and become a brand worth buying. Not "win the shelf, then hope for demand." Build the demand, and the shelf becomes a formality.

The part the victory lap leaves out

If this piece stopped here it would be a victory lap, and victory laps make bad advisors. This road has a cost, and you should see the invoice before you decide to walk it.

The same machine that made Poppi famous also made a claim the evidence couldn't fully carry. The cans told people, in effect, that the drink was good for your gut. A class action argued that the amount of prebiotic fiber in a single can was too small to deliver the promised benefit — and that you'd have to drink several a day, sugar and all, to get much of anything. The brand's former owner, VNGR Beverage, settled for $8.9 million. It admitted no wrongdoing, and the settlement received final court approval in April 2026.

Read that carefully, because the lesson is precise: in a better-for-you brand, a health claim that outruns your proof is not a free marketing lever. It's a line item on a future balance sheet. Building demand and building belief are close cousins, and the second one has legal edges.

There's a smaller cost worth naming too. Spending to create demand isn't the same as spending well. Poppi's 2025 Super Bowl push — custom vending machines sent to influencers — drew real criticism as wasteful. Poppi was big enough by then to absorb it. A brand your size might not be. This road quietly assumes you can tell the difference between spend that builds a habit and spend that just builds a headline. Not everyone can, and the bill for guessing wrong comes due fast.

When most of these are false, this is the road that ends in a quiet removal from the shelf and an expensive lesson — not a billion-dollar exit. The visible moves are the cheapest part to copy and the most dangerous part to copy first.

There's more than one road — and that's the relief

There's a cleaner way to see all of this, and it's sitting right next door in this series.

Grüns also sold into PepsiCo's world of better-for-you brands. Same era, roughly the same billion-dollar outcome. But Grüns took the opposite road. It didn't create a new category or manufacture a cultural moment. It owned an existing occasion — the daily nutrition ritual — with focus and restraint, spending on very little and refusing most of what it could have chased. (We wrote about it in Grüns Reached a Billion-Dollar Exit in Three Years.)

Two roads. One brand created demand and spent to spread it. The other captured demand that was already there and guarded its focus. Both ended up worth buying. Which tells you something that should be a relief: there's more than one way to build a brand someone wants.

"There's more than one road to a brand worth buying. The mistake is copying the road that fit someone else's product — and paying for a discipline you don't actually have."

So — do you need to go viral?

Back to the thought you had while scrolling. Do I need to go viral?

No. You need to decide which road your product is actually on. Capturing demand and creating it are both real, and both have produced billion-dollar exits. But they're different disciplines with different bills. One rewards restraint and focus. The other rewards nerve, capital, and airtight claims. The failure isn't picking the "wrong" one. The failure is copying the road that fit someone else's product — the Super Bowl ad, the viral stunt — and quietly paying for a discipline you were never set up to run.

The two decisions that made Poppi were made on purpose, years before anyone was watching. Yours can be too.

Related from Sound Decisions: Grüns Reached a Billion-Dollar Exit in Three Years · Distribution Doesn't Create Demand; It Captures It · Can Your Mission Survive Acquisition? · Purely Elizabeth Took Eight Years to Raise $3 Million

They’ll Pay for the Work. They Won’t Pay for the Thinking.

Strategy goes on the estimate, and strategy comes off the estimate. The usual answer is to argue that strategy is valuable. The numbers point somewhere less comfortable: this is a packaging problem, and the package costs more than the line was ever worth.

Every agency owner has had this meeting. The client approves the production, approves the media, approves the retainer, and then asks what exactly the strategy line is for. You explain. They nod. The line comes off.

Owners describe it almost word for word: we see the need, but the client won't pay for it. It gets said with a shrug, because it feels like a fact about clients. It is a fact. It just isn't a fact about clients. It is a fact about how the thinking was sold.

Put anything on its own line of an estimate and you have made it optional. That is what separate lines are for: they let someone see the parts, and seeing the parts lets them remove one.

Now ask which part is easiest to remove. Production defends itself. Take out the film and there is no film. Strategy is the one line a client cannot judge until after they have bought it, and its absence shows up eighteen months later as a campaign that ran fine and a business that didn't move.

The people reading the estimate are paid to shorten it. In the ANA and 4A's 2023 study of agency reviews, cost and price was the top factor clients used to pick a winning agency, named by 62 percent, and procurement staff weighted it highest of all. An itemized strategy line walks straight into that. The sequence is not the client doesn't value thinking, so they cut it. It is the agency itemized the thinking, itemized things get compared, and the hardest one to judge loses. You cannot change how clients value judgment. You can change a document.

Hold everything constant except the position the agency holds when the work starts. Same brief, same acquisition cost. In one version the agency is a vendor with a scope. In the other it is the agency of record, trusted with the calls about the client's business.

Nothing about the client changed. The seat did. So when a client crosses out the strategy line, the number they think they are saving is a fraction of a scope. The number actually moving is the lifetime value of the account, and it moves the wrong way for both of you.

A vendor holds a scope: finite, easy to describe, easy to compare, and therefore put back out to bid on a schedule. Being reviewed is not a verdict on the work. It comes with the position. An agency of record holds a seat, and seats are renewed rather than re-bid, because replacing one costs the client the judgment that came with it.

The tenure data shows the difference plainly. In April 2025 the ANA and the 4A's reported that agency-of-record relationships now last about seven years on average. Where the client has no mandatory review period, which is 60 percent of clients, they last 8.1 years. Where reviews are frequent, as little as 3.8.

What moves an agency from scope to seat is not better production; the vendor already produces well, which is why it was hired. It is the point at which the agency starts making calls the client cannot make alone, about the business rather than the brief. That is what strategy is. It survives badly as a line and well as the basis of the whole relationship.

The same ANA and 4A's study found independent agencies keep agency-of-record clients 7.3 years on average, against 5.8 for holding-company agencies. The independents win that comparison against firms with strategy departments and whole floors of people titled planner.

The likeliest reason is proximity. At an independent, the owner is on the account, and the person who can see the client's business is the same person answering their email. The holding companies sell that judgment at a rate, with a department behind it. Independents tend to give it away: on calls, in hallway conversations, in the Sunday-night note that never reaches an invoice.

The market has moved toward what they are giving away. Promethean's 2026 guide describes the pitch that wins as having shifted from execution capability in 2021 to judgment, integration and outcomes in 2026. Yet account management, the function that sits closest to the seat, makes up about 6 percent of a typical agency's headcount, against roughly 63 percent in production.

The eight-to-one figure is arithmetic. It scales one published case by a published tenure average. It shows the size of the stake, not what any particular agency will earn.

The tenure figures are self-reported by marketers and agencies answering a trade-body survey, and they describe agency-of-record relationships specifically.

Promethean sells growth reviews to the agencies it studies. Its figures agree with the independent tenure data in direction; weigh the exact numbers accordingly.

None of this works if the thinking isn't there. If what sits under the strategy heading is a trend deck assembled the week before the pitch, the client who declines it is being accurate, and no change of packaging fixes that. One test settles it: if your senior people left tomorrow, would the client notice a loss of judgment, or only a loss of output?

Senior judgment does not scale by hiring. Un-itemizing works on the accounts the owner personally touches. The rest of the book is a harder problem, and adding senior headcount is the usual answer precisely because it is the one that squeezes margin.

A client has never declined judgment. They decline a line item called strategy. Those are different transactions, and the gap between them is worth about four years of tenure on a single account.

Related from Sound Decisions: Your Next Client Is Already Inside This One · Nobody Pays You Extra for Being Honest · The Marketing Job Is Derived Power

You Don’t Need More Clients. You Need Longer Ones.

The industry's default route to growth is more people and more logos. On its own published numbers, that route buys revenue and gives up profit. The trade almost nobody is running is the one on the clients they already have.

Nothing here will tell an agency owner to want less. If you want to grow, grow. The question is which kind of growth, because the industry's default answer, hire more people and win more logos, turns out on the industry's own numbers to be the most expensive road available.

Start with what happens to profit as agencies get bigger.

That is a comparison between firms of different sizes, not a record of the same firms growing, so it does not prove that growth halves any one agency's margin. It does show where the size curve points. And the same research carries the part that spoils the simple story: larger agencies grew fastest in 2025 and small ones grew slowest. Headcount is not a con. It buys revenue.

Headcount buys revenue. Tenure buys profit. Most owners believe they are buying the second when they are buying the first.

Among agencies that track it, Promethean found average project margin, the profit on the work itself, was 35 percent. Average net margin, what reaches the owner, was 13. Twenty-two points disappear between a job going well and the money arriving. They go to coordination, management layers, delivery oversight, reporting, recruiting and the systems that hold it together, which is exactly the layer that headcount growth multiplies. Only 59 percent of agencies measure project margin at all.

Now look at where agency people sit. Promethean analyzed 3,172 positions across 1,228 agencies. Production is about 63 percent of headcount. Account management is about 6 percent, and sales and marketing about 7. Set that against the same firm's finding that growing, upselling and keeping existing accounts produces more revenue per unit of effort than any acquisition channel it measures.

Most of the payroll sits in the layer whose price is falling, and the layer that decides whether the biggest client is still there in three years is, at most independents, one person: the owner.

If you have privately suspected this is a personal failing, put that down. Promethean's own description of how agencies organize for revenue shows it at every size: the revenue function grows from a solo founder at the smallest shops to a full revenue leadership team at the largest, and even at the top, partners still come in to close large, transformative accounts. The judgment that wins and keeps the biggest relationships stays with leadership, because very few firms have ever built a system that works without it.

That is a structural property of the industry, and it has now been measured. It also means the obvious fix, hiring senior people to carry that judgment, is the one that pushes a firm down the size curve toward eight cents on the dollar.

If tenure is where the profit is, the thing that decides tenure matters most. In the ANA and 4A's 2024 follow-up study on agency reviews, 90 percent of clients said the overall value and long-term return of the work outweighed cost when engaging an agency. Then came the gap: only 5 percent of clients, and 10 percent of agencies, had a formal, corporate-backed definition of what value means.

Nine in ten clients buy on something that almost none of them have written down. Reviews get triggered on it and contracts get lost on it, and for nearly everyone it exists as a feeling. An agency that writes it down with its biggest client, in that client's numbers, has done something roughly nine in ten of its competitors have not.

Specializing will not close this gap by itself. Promethean reports that 86 percent of agencies now call themselves specialists, up from about 30 percent a decade ago, and that agencies which cut services grew fastest in 2025 and averaged 30 percent net margins. Focus is a real margin tool. It changes what you sell. It does not change the position you hold in the client's business.

The size comparison is cross-sectional. Small and large agencies may differ in more than size: services, clients, ownership. The 19-to-8 gap describes the industry, not a forecast for your firm.

The margin data comes from digital agencies. Integrated and creative shops may sit elsewhere on the curve.

Promethean sells growth reviews to the agencies it studies, and its return ranking is a ranking, not a measured percentage.

Longer clients are not fewer clients. Deepening a book without adding to it builds dependency. The working rule in agency finance, a practice convention rather than a survey finding, is that a client consistently above about a fifth of revenue has stopped being the best account and started being the board. New business is what makes deepening safe.

Growth that only replaces churn is a treadmill you are paying to run on. The agencies that keep more of each dollar are not the ones that hired fastest. They are the ones whose clients stayed long enough for the work to pay.

Related from Sound Decisions: They’ll Pay for the Work. They Won’t Pay for the Thinking. · Your Best Clients Are the Ones You Undercharged · Your Next Client Is Already Inside This One

Where AI Actually Pays in an Agency. The skeptic was half right.

The most quoted AI statistic says 95 percent of pilots returned nothing. The findings underneath it are more useful: what the paying minority bought, how they measured it, and the risk most agencies are already carrying without having bought anything at all.

The invoice arrives every month now: the seats, a couple of specialist tools, an enterprise plan someone signed up for in the spring. The team says the tools save time, and you believe them. But nobody can point to where the saving shows up in the numbers, and at some point in a partners' meeting someone asks it out loud: what are we actually getting for this?

If you have been the one asking, this piece is written for you. And the most-quoted number in the whole debate is on your side.

In 2025, MIT's NANDA initiative reviewed more than 300 public enterprise AI deployments, interviewed 52 organizations and surveyed 153 leaders, and reported that 95 percent of generative AI pilots produced no measurable impact on profit and loss, against $30 to $40 billion of spending.

That headline deserves the same scrutiny you would give any vendor's claim. The report is industry research, not peer-reviewed; its full dataset has not been published; and "no measurable impact" is not the same finding as "failed." Critics have argued that much of what it records is spending with no baseline to measure against. Read that way, the report is less an obituary for AI than a description of how most of it was bought.

The more useful findings sit under the headline. Tools bought from specialized outside vendors succeeded about twice as often as systems built in-house. Budgets leaned heavily toward sales and marketing tools, while the report found better returns in narrower back-office work. The report's own explanation for most of the gap was organizational rather than technical: tools that did not fit or learn the workflow they were dropped into. The practical corollary is to measure a task's cost before the tool arrives, because without that no return can be proven either way.

Each of those is a capital-allocation decision, the kind a finance lead already knows how to make. None of them is a bet on the technology.

Your people adopted AI a year ago, whether or not the firm did. The open question for most agencies is not whether to use it but whether its use is governed. For a business that handles other companies' plans, launch dates and sometimes customer data, ungoverned use is an unpriced liability, and it stays unpriced until one client asks where their material went.

The clients are also moving. In Gartner's 2025 CMO Spend Survey of 402 marketing leaders, 39 percent said they planned to cut agency budgets, and 22 percent said generative AI had already reduced their reliance on external agencies for creativity and strategy. The work AI makes cheapest is the work clients can increasingly do without you.

Apply the four questions to how most agencies use AI and a pattern shows up: the spending sits in drafting, variations and first-pass audits, the layer whose price is falling fastest. The companion piece on this site, on why AI multiplies the process a firm already has, covers the other half of the problem.

The MIT NANDA figures are not peer-reviewed, and the dataset is unpublished. The initiative also studies agent infrastructure, so it has a point of view about what the answer looks like. Treat the 95 percent as direction, not measurement.

None of the samples are agencies specifically. The enterprise findings describe large companies; how exactly they carry to a thirty-person shop is a judgment.

Gartner's figures are from spring 2025 and describe intentions, not completed cuts.

"Bought beats built" is not a rule for every firm. A specialist vendor is still a dependency. A firm with a genuinely unusual process may have to build, and should do it knowing the base rate.

The skeptic was mostly right about the headline and mostly wrong about the conclusion. AI does not reliably pay. The firms it pays for bought narrowly, measured first, and knew where their clients' material was going.

Related from Sound Decisions: AI Doesn’t Give You Leverage. It Multiplies the Process You Already Have. · Creative Is Now a Line Item · You Don’t Need More Clients. You Need Longer Ones.

The Market You Proved. A $4.4 trillion economy, a 4% budget.

The case that the U.S. Latino market matters was made decades ago and won. The budgets never followed: brands still plan under four percent of advertising for a fifth of the country. For specialist agencies, the next fight is not proving the market. It is collecting on it.

Forty years ago, the founders of America's Hispanic agencies had to argue that their market existed. They made the case in rooms that did not believe them, with budgets that did not match the opportunity: that tens of millions of Americans were being spoken past, that culture was not a translation problem, that the growth was real.

They won that argument. The institutions they helped build now publish the numbers the whole industry cites. What did not follow the proof was the money.

The economy is not only large; it is where growth is coming from. The same report puts Latino household consumption at $2.8 trillion in 2024, and in Texas alone the Latino economy reached $820 billion. The Hispanic Marketing Council has a name for the cosmetic response that often stands in for investment: "Latino coating," a campaign with Latino elements on the surface and nothing in the plan.

The dedicated budget did not disappear. It was routed elsewhere, over roughly fifteen years, down several paths at once.

The competition a specialist agency faces is rarely more agencies like it. It is everyone else: general-market shops, in-house teams, networks and platforms, all claiming the territory the specialists proved valuable, while the dedicated budget sits under four percent.

A wall of iconic campaigns earns respect in every room and wins fewer of them each year, because respect is not what chief marketing officers are buying. They are buying growth they can attribute, and every one of the four exits claims it can deliver that more cheaply.

What the exits cannot copy is judgment: knowing this market as a set of communities rather than a segment, which insights are real and which are coating, what earns trust and what burns it. That is the one asset in this story that gets more valuable with age. But it converts to revenue only in rooms, the planning meeting where the total-market brief gets written and nobody reopens the four percent, and only when it is argued in the buyer's currency: growth, attribution, share of the category's gains.

"Under 4%" measures targeted spend, not reach. General-market campaigns also reach Latino consumers. The figure shows how little is planned specifically for them, not how little reaches them.

The budget figure dates from 2024, as cited by the Hispanic Marketing Council from media-spend data; the economic figures are from 2024 and published in 2026.

The Latino Donor Collaborative is an advocacy think tank, with research produced by Arizona State University's business school. Its estimates are the most widely cited available and are built from federal data; they are estimates, not official national accounts.

This is not a case that every brand should run a dedicated program. For some categories and audiences, well-built total-market work is the right answer. The argument is that the decision should be made on the numbers, in the room, rather than by default.

The argument that the market exists was won a long time ago. The collection happens the same way the argument was won: one room at a time, now in the buyer's language rather than the market's.

Related from Sound Decisions: Anyone Can Look Like You Now · Your Best Clients Are the Ones You Undercharged · How to Win Houston's Growing CPG Market

Houston Is Booming. Will Your Agency Boom Too?

Houston added nearly 127,000 residents in a year and ranked third among major metros for job growth. Its professional-services firms had a harder 2025, and the pressure that outlasts the dip is pricing: costs rising faster than rates, while AI shrinks the hours an agency bills.

It is renewal season in a good year. The phones are busy, the team is stretched, and the city outside is growing faster than almost anywhere in the country. You send your biggest client a rate increase of a few points and feel reasonable about it. The same week, your best strategist asks for a raise bigger than that. And someone on the team mentions, as good news, that the monthly reporting now takes a fraction of the time it used to, because the AI tools do the first draft.

Each of those things feels fine on its own. Together they describe a firm getting busier and poorer at the same time, and a booming city makes that very easy to miss.

The boom is real. The Greater Houston Partnership reports the metro added about 126,700 residents between July 2024 and July 2025, and its real output grew 10.6 percent from 2022 to 2024, second among the twenty largest U.S. metros. In its midyear update this August, the Partnership counted 39,700 jobs added over the past year, third among major metros.

And yet 2025 was not kind to the firms that sell expertise to other businesses. When the Partnership broke down the year, it found six sectors lost jobs, concentrated in office-heavy industries, especially professional, scientific and technical services and administrative support. Its explanation: softer client demand and tighter discretionary spending, partly from upstream oil companies cutting back on services and consulting as energy prices fell. The town boomed. Spending on firms like yours did not keep pace.

Hiring has since picked up across Texas, and professional and business services grew statewide in the second quarter of 2026. So the question for an owner is not whether demand comes back. It is narrower and more durable: when it does, can the way you charge keep up with what you pay?

For most businesses that gap is a squeeze. For an agency that bills by the hour, it is worse, because of a second pressure running the other way. AI makes the same work take fewer hours, and an hourly firm bills fewer hours for it. Every hour the tools save is an hour that can no longer be charged, while the salaries of the people using those tools rise about four percent a year.

One respondent to the Dallas Fed's March 2026 survey put the general problem in a sentence: labor and input costs keep rising, but there is little to no room to raise selling prices. An agency adds to that the fact that its unit of sale is shrinking.

The other pressures on Houston's independent firms are covered elsewhere in this series: the referral channel thinning as clients turn to search (The Budget Line), the cost and odds of the open pitch (Your Next Client Is Already Inside This One), and the execution work moving in-house (How to Win Houston's Growing CPG Market). The pricing model is the one pressure an owner controls directly.

The Dallas Fed figures are Texas-wide and cross-industry, not specific to agencies or to Houston. Service-sector wage growth alone was 3.8 percent in June.

The 2025 job losses were a year, not a trend. Houston hiring has recovered in 2026, and the Partnership expected professional and technical services to add jobs this year. The squeeze on pricing is the more persistent problem than the dip in demand.

Fixed fees carry their own risk. A firm that scopes badly loses money on every job it prices as a result instead of as time. Moving off hours requires knowing your costs per job better than most firms do.

How much AI cuts hours depends on the work. Production and first drafts compress more than client strategy does.

Houston's clients will keep growing. Whether the firms that serve them grow too depends less on the economy than on whether they are still selling the one thing getting cheaper every quarter: their hours.

Related from Sound Decisions: They’ll Pay for the Work. They Won’t Pay for the Thinking. · AI Doesn’t Give You Leverage. · You Don’t Need More Clients. You Need Longer Ones.

How to Win Houston's Growing CPG Market. The shelf is recruiting. The platforms are self-serve.

Houston's consumer brands are forming in plain sight, recruited by the state's biggest grocer and handed self-serve tools to run their own shelf campaigns. What is left for an outside firm is the part no platform supplies: knowing which few moves change velocity.

Look at the client list of most independent agencies in Houston and you will see the city's economy reflected back: energy, healthcare, industrial, professional services. Consumer brands are something that happens somewhere else. The famous Texas better-for-you exits cluster in Austin, and when an owner here thinks about food and drink work, it usually sounds like a pitch they would have to travel for.

That picture is out of date. The brands forming in Houston are not famous yet, which is exactly why they need help, and the state's biggest grocer is already out looking for them. The question for an agency owner is not whether that client pool exists. It is whether your firm is the kind those founders will need.

The Greater Houston Partnership reports the region led every U.S. metro in manufacturing output for the third year running, $126.9 billion in 2024, ahead of Los Angeles and Chicago. That figure is dominated by petrochemicals and energy; it is the terrain, not the crop. What matters for a food or drink brand is what that industry left behind: the port, the cold chain, the co-packers and the logistics that a young consumer brand needs. The same report counted a record 683 new business announcements in 2025, up 26.5 percent.

What makes the client pool real rather than theoretical is that the retailer these brands need most runs a standing program to find them. H-E-B's Quest for Texas Best, now in its thirteenth year, has reviewed more than 7,000 product samples and brought more than 1,000 Texas-made products to its shelves. This year's ten finalists were announced in September and pitch in San Antonio on October 8; three are from the Houston area. In 2025, four Houston companies made the top ten.

Once a brand is on the shelf, the next game is retail media. H-E-B Retail Media reaches more than eight million households a week, and in June 2025 it launched self-service tools so brands can run and measure their own promoted-product campaigns on heb.com and the My H-E-B app. Its general manager framed the goal as making it easier for brands of all sizes to reach local Texas communities.

The work agencies lead with is the work that moves inside first. The ANA reported in 2023 that 82 percent of its members run an in-house agency, up from 42 percent in 2008, and that 65 percent had moved some work from outside agencies to their own teams in the prior three years, led by social, search, email and other digital creative. That is a national figure among larger brands, and the picture has since become mixed: in RSW/US's 2026 outlook, clients reported less of their work handled in-house than in 2024, even as agencies reported more clients building some in-house capability. The direction for execution work is still clear.

The tools are built to be used without an agency. A self-service retail media platform is exactly that: a founder can log in, set bids and read results without anyone in the middle. None of this means the agency is finished. It means one version of the agency is, the one whose value is operating tools a brand can now operate itself.

There is no famous CPG agency in Houston yet, which is the opening. The first firm to out-prepare a founder on the shelf, slightly ahead of its own case studies, and then do the work, becomes that agency. The case study follows the client, not the other way round.

Manufacturing output is not a count of consumer brands. The $126.9 billion is all manufacturing, mostly heavy industry. It establishes the infrastructure, not the size of the client pool.

Quest is statewide. The 55 million-dollar brands are Texas-wide; the Houston share of them is not published.

The in-house data is national and from larger brands, and the latest client-side reading suggests some work is moving back out. Young founder-led brands may behave differently from ANA members.

A small brand is a small budget. Winning a founder on the way up is a bet on their growth; the first engagement may be modest and the payoff years away.

The clients are forming in Houston's own backyard, funded, recruited by the state's biggest grocer, and short on the one thing a platform cannot give them. The firm that understands their shelf before they ask will be the one they call.

Related from Sound Decisions: Your Velocity Number Means Three Different Things · Why Whole Foods Is Not the Holy Grail · Anyone Can Look Like You Now

You Caught the Protein Wave. So Did Everyone.

Putting protein on the front of the pack feels like catching up to where the market is going. But a claim every brand on the shelf is making has stopped selling anything — and the growth was never in the attribute.

You're looking at the mockup of your new pack, and it finally looks right. 20G PROTEIN across the front, a callout about fiber, a line about being made for the way people eat now. For the first time in months it feels like you've caught up — like the brand is finally pointed where the whole market is going. It's a good feeling. It's also the feeling right before the problem: that same quarter, a dozen brands bigger than yours are printing the exact same words on their own packs, and the shopper standing in the aisle is about to see all of you at once.

The wave is real

Start with what's true, because it's why you added the claim in the first place. The move toward protein isn't a fad you can wait out — it's the shape of how people eat now. About one in eight U.S. adults now takes a GLP-1 drug like Ozempic, per KFF polling, and those households shift their spending toward protein, fiber, and fresh produce and away from sweet and salty snacks. It runs well past the people on the drugs: nearly all of Gen Z — 98%, by Mintel's count — has eaten a functional food or drink in the past six months, with protein and fiber the headline benefits. If you sell better-for-you food and you weren't thinking about protein, you'd be missing where demand actually is. Catching the wave was the right instinct. The trouble is what a wave becomes when everyone catches it at once.

The wave is also a flood

Protein is now the front-of-pack flag on cereal, coffee, soda, ice cream, pasta, even candy — categories that had nothing to do with it a few years ago. And Mintel's 2026 read names the cost plainly: the flood of protein and fiber claims is blurring the lines between categories and slowing the very growth it was meant to create.

Here's the part most founders miss. A claim every brand can make isn't a differentiator — it's the cost of admission. When your 20g protein sits beside nine other packs saying the same thing, the words stop doing any work. You can be exactly on-trend and completely invisible at the same time. That's the velocity wall in a single picture: you did the on-trend thing, and it moved nothing off the shelf — because on-trend is now the baseline, not the edge.

"An easy way to add 15 grams of protein to virtually any cold beverage."

Brian Niccol, Starbucks CEO, on the chain's new protein cold foam

When a differentiator is something a giant can sprinkle onto anything, it was never a differentiator.

We know how this movie ends, because it already played once — with the exact same attribute.

What holds is the occasion, not the ingredient

This is the whole game — and the line between a decision that moves velocity and marketing that only keeps up. A durable brand owns a moment a shopper reaches for; a fad owns an ingredient that ages out. Protein is an ingredient. The 3 p.m. crash. The thing my kid will actually eat. The post-workout ritual. Dessert I don't have to feel guilty about. Those are moments — and a moment can be owned in a way a nutrient never can. The question that moves product off the shelf was never "how many grams can I claim?" It's "what moment do people reach for me by name to solve?" Answer that, and protein becomes a feature of the thing you own. Skip it, and protein is just the flag you're waving next to everyone else's.

Two calls, both about what not to spend on

First, don't spend your one scarce differentiation budget on the attribute everyone already has. Clear it quietly, as table stakes — then win where you're alone on the shelf: the occasion, the identity, the reason someone chooses you specifically.

Second, don't build the brand on a cohort that leaves. It's tempting to make GLP-1 friendly the whole identity. Be careful — per AlixPartners, more than half of GLP-1 users stop within about a year, and their grocery baskets drift back toward what they bought before. Protein-for-satiety belongs inside your occasion as a feature, not as the reason the brand exists. And the treadmill is real: Mintel is already calling fiber the next must-have nutrient. Chase the attribute and you re-run this whole exercise every eighteen months, always a step behind the flood.

The wave isn't going to stop. Protein now, fiber next, something after fiber the year after that. Catching each one faster than the last is a treadmill that never pays off — the moment you're on it, so is everyone else. The brands that last aren't the ones who catch the wave first; they're the ones who own a stretch of shore the waves keep washing up on. Protein isn't your brand. The moment someone reaches for you by name is. Everything else is a claim — and claims are cheap now.

Getting Into Walmart Is the Dream. It Can Also End the Company.

Landing Walmart feels like you've arrived. But the account big enough to make your year is big enough to remake your business and then strand it — and this is the rare danger that comes from winning, not failing.

You'll remember where you were when the call came. Maybe it was a buyer, maybe your broker, maybe an email with a subject line you read three times to be sure. Walmart wants the line — not a test in forty stores, a real rollout, hundreds of doors, maybe more. You said the number out loud to whoever was nearest. Someone opened something. For a minute, the years of samples and rejections and almost-theres all paid off at once, because this is the thing every founder is told to want: the biggest shelf in the country saying yes.

Hold onto that feeling, because the rest of this is going to argue with it. Not because the yes is bad. Because of what it quietly starts.

There's a version of the shelf conversation that's about whether your product will sell — whether you have the velocity to hold the slot. This isn't that one. Assume it sells. Assume Walmart works. The danger I want you to see is the opposite of failing: it's what winning, on Walmart's terms, can do to a business that was already sound.

The account that rebuilds the company

Here's what it took me years inside big food to watch happen, over and over, to good brands run by smart people. The biggest customer doesn't just buy from your business. It starts to rebuild it.

To serve one account that size, everything quietly bends toward it. You add a production line for their volume. You change your pack sizes to their spec. You take on their delivery windows, their paperwork, their terms. You front the cash to make all of it months before you're paid for any of it. Your best people spend their weeks on this one relationship, because it's now too big to get wrong. Decision by decision — each one reasonable on its own — you stop being the company you built and become a supplier to one customer.

None of those choices feels like a mistake in the moment. Together, they add up to a different, more fragile company than the one that walked in.

"The account big enough to make your year is big enough to remake it."

One crop, all the power

Now look at who holds the leverage. The bigger that one account grows, the less say you have in it. Walmart's whole promise to its shoppers is the lowest price, every day — which means the ask, year after year, is a lower price from you. And when most of your volume runs through one buyer, you have no real counter. You can't walk away. They can.

The most famous version of this is a jar of pickles.

There's a second cost, quieter than price. A business that leans on one customer is simply a more fragile business — and everyone who might lend to you or buy you someday knows it.

And the cash moves the wrong way as you grow. You're fronting production for that volume and waiting to be paid; meanwhile a late or short shipment gets docked straight off your invoice — Walmart runs a program that takes a few percent of the cost on deliveries that miss the window, and suppliers learn to budget for it. So the bigger the account gets, the more of your cash is locked inside it, and the less of it is yours to steer. Growth that starves your cash can sink a business that looks, on paper, like it's winning.

A field of one crop

There's a reason this pattern has a name in farming. The world's bananas were once a single variety, the Gros Michel — grown everywhere because it shipped well and sold big. One soil-borne disease swept through and wiped out the global crop. The banana we eat today, the Cavendish, is now failing to a new strain — because it, too, is a single crop, with no variety to fall back on.

A field planted with one crop has no defense against the one thing that crop can't survive. That's the real risk of the big account. Not that it fails you — that you've bent your whole field around it, so when it turns (a delisting, a renegotiation, a category that moves), there's nothing else left standing. And the tempting escape — swap one giant customer for a different giant customer — isn't diversification. It's the same bet with a new name.

We call this The Monocrop — it has its own page, with a 90-second check on how exposed your own field already is.

When Walmart is the right call

None of this means Walmart is a mistake. For the right brand, at the right time, it's one of the best things that can happen to a business — enormous reach, real volume, a genuine partner. The brands it rewards tend to share a few things, and they're worth being honest with yourself about before you say yes.

If those are true, go — and go with your eyes open. If they're not, the answer isn't no forever. It's not yet, and not like this: start smaller, keep your pricing power, and don't let one yes remake the company before it's ready to be remade.

The question underneath the celebration

So before the toast, ask the question the toast is sitting on top of: which of my customers is big enough to end me — and what am I doing so the answer is none of them?

Plant for variety. Keep enough balanced rows — accounts, channels, ways to be bought — that losing any single one of them is a bad quarter, not the end. That isn't caution; it's what keeps the field standing. The biggest yes you'll ever get is worth having. Just don't let it become the only crop you grow. Because the account big enough to make your year is the account big enough to end it — and which one it becomes is still, today, your decision to make.

When the Money Changes Its Mind

The money that once rewarded growth at any cost just changed its mind — and a generation of brands built for the old rules is feeling it. The move now isn't to spend less; it's to cut what can't prove it works, and put everything behind what can.

You know the meeting. A year ago, the people who wrote you a check wanted one thing: growth. More doors, more SKUs, more states on the map, more spend behind all of it. So you delivered — because that was the job, and because every signal in the market said the brand that grows fastest wins. Now the same people are in the same room asking a question that sounds almost like its opposite. Not how fast are you growing. How much does it cost you to grow — and what happens if we stop paying for it.

Nothing about your brand changed between those two meetings. The money changed its mind.

The rules repriced while you were busy growing

For most of the last decade, capital was cheap and patient, and it rewarded one thing above all others: top-line growth. You could spend ahead of demand, buy your way into doors, and trust that the next round would cover the gap. That era is over. Investors now describe a market that has swung from "grow at any cost" to "prove it pays," and a generation of brands that raised easily in the good years are finding themselves stranded between seed and Series A — not because the product got worse, but because the rule for getting funded got rewritten underneath them.

This is the part worth being honest about: a repricing is not a storm you wait out. It doesn't pass. It resets the terms you'll operate under for years. The founders who come through it well are the ones who stop treating the tight market as weather and start treating it as the new climate — and adjust what they're building accordingly.

"Spend less" is the wrong instruction

When money tightens, the reflex is across-the-board restraint: trim every line ten percent, slow everything down, wait for clarity. That feels responsible. It isn't a decision — it's a flinch. A flat cut starves your winners at exactly the same rate it starves your mistakes, which means you come out the other side smaller and no sharper than you went in.

The move that actually works is the opposite of uniform. It's a cut that sorts — that tells the difference between the spend earning its keep and the spend you're funding on a story about where it'll get to someday. Every dollar, every SKU, every door is doing one of those two things. In a market that stopped paying for hope, only the first kind survives contact with a board.

Feed what's proven. Cut what's hope.

The gate is velocity — does the thing actually move where you already are, fast enough to keep its slot? That's the number a retailer watches and the one a buyer watches, and it's the only honest test of whether a dollar is working. Distribution captures demand; it doesn't create it — so the doors you've won beyond your proven region, the packaging refresh while velocity is flat, the paid acquisition running ahead of your repeat rate: those aren't growth. They're hope, wearing growth's clothes. By one widely cited industry estimate, roughly four in five premium food brands never clear a million dollars in retail — almost always because they funded the hope and ran out of room before the proof arrived.

If this sounds severe, look up. The biggest player in the category is doing exactly this.

The cut protects the feed

A pruned brand is not a smaller brand. It's the same capital, aimed at the things that actually compound — which is what growth was supposed to mean before cheap money let everyone skip the step. And this is where the two things every health-forward founder is trying to hold at once stop being a tradeoff. You do not cut the mission. You cut the spend that was never moving it. The discipline that keeps you solvent is the same discipline that keeps the mission alive long enough to matter — which is its own decision, made years earlier than most founders think. Mission and momentum aren't opposing levers. The cut is how you get to pull both.

The cautionary version of this is written in the largest letters in the industry.

"Cutting what isn't working isn't retreat. In a market that stopped paying for growth, the disciplined cut is the only growth still on the table."

The Pruning Cut · CULT+MATH

The instinct in a downturn is to do more — more launches, more channels, more noise — to outrun the fear. The harder and better move is to do less, on purpose, and be able to say exactly why each thing that survives the cut is still there. That's not playing small. It's the only version of ambition the market is currently funding.

And it's a CEO's call, not a marketer's — which matters, because a tight year is precisely when founders get pulled back into running the spreadsheet instead of running the company. Make the cut deliberately, from the front, and you get the seat back.

Related from Sound Decisions: Purely Elizabeth Took Eight Years to Raise $3 Million · Can Your Mission Survive Acquisition? · Poppi Didn't Go Viral by Accident

Distribution Doesn't Create Demand; It Captures It

Winning more shelves feels like growth. It isn't — distribution can only capture demand you've already built, and expanding ahead of it just starts more clocks you can't beat.

You know the email. It lands on an ordinary Tuesday, from a buyer or from your broker, and the tone is friendlier than the message underneath it: the category is going into review, your product is selling below the number the shelf needs, and you have a few weeks. You read it twice. You think about the months you spent getting into that chain — the samples, the forms, the meetings that went nowhere before the one that finally didn't — and how quickly all of it is suddenly in question.

Then your mind does the thing almost every founder's mind does. It starts looking for more shelves. Another chain. A bigger banner. A regional rollout to prove the brand has momentum. If the problem is that the product isn't moving, surely the answer is to get it in front of more people.

That instinct feels like growth. It's the trap.

Shelf space is the doorway, not the win

Here is the thing it took me years inside big food to see plainly, and that most founders learn the expensive way: shelf space looks like the win, but it's only the doorway. Distribution doesn't create demand. It captures it.

Getting onto a shelf does exactly one thing — it puts your product within reach of someone who might want it. That's not nothing; people can't buy what they can't find. But reach is not desire. A shelf can hold your jar at eye level in a thousand stores and not move one of them any faster if the want isn't already there.

"Distribution is a net. It can only catch the fish that are already swimming."

Ipalibo Da-Wariboko, CULT+MATH

Every shelf starts a clock

The net has a cost. The day you win a slot, a quiet clock starts running. The buyer has a number in mind: how fast your product needs to sell, in each store, each week. That rate is your velocity, and it is the single number that decides whether you keep the slot or lose it. Miss it for long enough and you don't get a warning — you get cut at the next review.

This is why store count can be the most dangerous vanity in your business. It's easy to feel successful shipping cases into a distributor and forget that cases shipped are not the same as product sold to a shopper — and only one of those pays you twice. Every new store you add without the demand to fill it doesn't spread your risk. It multiplies it. Now you're paying to sit on more shelves, each one running its own clock, each one quietly counting down. Expand faster than your demand and all you've done is speed up the moment the shelf takes itself back.

Demand is made before the shelf

So if distribution can't create demand, what does?

The demand was always made upstream — long before the shelf, in three things that have nothing to do with how many doors you're in.

The first is the occasion you own: the specific moment in someone's day or week when they reach for your product on purpose, by name. Not "a healthy snack" — the thing you grab at three in the afternoon when the alternative is the vending machine. A brand that owns a moment gets bought without being deliberated over.

The second is a product worth buying again. Trial is cheap; almost anything gets tried once with the right promotion. Repeat is the whole game. If people don't come back on their own, no amount of distribution fixes it — it only exposes the problem faster, across more stores.

The third is being remembered. Enough room in your shopper's mind that when the occasion arrives, you're who they think of before they're standing at the shelf deciding.

You can see all of this in who survives — and who doesn't.

The brands that grew the other way look different. Olipop spent its early life proving, in a small handful of stores, that people came back for the soda — before it scaled. Liquid Death built a crowd that sought the can out before it ever chased the national shelf. The same lesson, twice: they created the demand first, then let distribution catch it.

Distribution still matters — but only in that order

None of this means distribution doesn't matter. It does. The people who have studied how brands actually grow are blunt about it: broad availability is necessary. You can't grow if people can't find you, and a great product nobody can reach is just a hobby. But necessary is not the same as enough. Availability gives you the chance at growth; it never guarantees it.

So the mistake is almost never caring about distribution. The mistake is the order — reaching for the next door before you've earned the demand to fill the door you already have. Getting onto the national shelf before your product moves isn't a shortcut past the hard part. It's the hard part, at a larger and more expensive scale.

The question to ask before the next door

Before you answer that Tuesday email by chasing more shelves, ask the harder question first: have I built the demand to fill the shelf I already have?

If the answer is yes, then go — distribution will compound everything you've built. If the answer is no, more stores won't save the brand. They'll only start more clocks. Distribution doesn't create demand. It captures it. The work that actually grows your brand — the occasion, the repeat, the place you hold in your shopper's mind — happens before the shelf, and it's the part you can still control.

Grüns Reached a Billion-Dollar Exit in Three Years. The Decisions Came First.

Unilever completed its acquisition of the greens-gummy brand on June 1, 2026, at an estimated $1.2 billion — terms undisclosed. The speed is the headline; the early decisions underneath it are the lesson.

If you're building a health-forward brand, you already know the feeling. You have the mission. You may even have the shelf. What you're watching, every week, is the one number that decides whether you keep it: how fast your product actually moves once it's there. And when a brand like Grüns goes from a standing start to a billion-dollar sale in three years, the temptation is to file it under luck, or timing, or a founder who knew the right people — anything that lets you keep doing what you're already doing.

That reading is comforting, and it's wrong in the way that matters. Grüns didn't win at the end, in the part everyone can see — the retail rollout, the celebrity investors, the acquisition. It won at the beginning, in a handful of decisions most founders skip because they don't feel urgent. By the time the growth showed up, the outcome was mostly already decided.

Here's what those decisions were — and, just as important, which of them you can actually copy.

First, the facts — stated carefully

Unilever completed the acquisition on June 1, 2026, and Grüns is now part of its Wellbeing business. The terms weren't disclosed; the widely-repeated ~$1.2 billion comes from Axios and hasn't been confirmed by the company. What's not in dispute is the shape of it: founder Chad Janis launched the product in August 2023 and signed the deal about 32 months later, having raised roughly $55 million in total along the way.

Sit with that for a second. That gap — between what Grüns spent and what it turned out to be worth — is not a growth story. It's a decision story.

The exit was won upstream

In this category, brands don't usually die from a weak mission. They die from weak velocity — product that doesn't sell off the shelf fast enough, per store, per week, to clear the retailer's hurdle. When the number is too low, the buyer doesn't call to talk about your values. They take the slot back. And velocity isn't won on the shelf. It's won upstream — in the position you take and the occasion you own, long before a single store carries you.