The question is registered before the data is pulled, and every row is published.
These are standing studies rather than articles. Each one names its question and what would make it false before any retrieval, states its method, publishes the rows behind every figure, carries its limits on its face, and is re-run on a schedule. When a measurement overturns an earlier one, the correction is published and dated rather than quietly applied.
Study data on these pages is published under CC BY 4.0: reuse it freely, crediting CULT+MATH and the study date. The paid Decision Briefs carry their own license.
Streaming
What happens to attention after the opening week.
AI
How the answer engines behave, and which vendors people reach for.
CPG
Margin, pass-through, and demand read from the filings.
Film
Formats, windows, and what an audience score measures.
Games
Dated launches, public conversation, published numbers.
Marketing
What companies actually spend, from their own filings.
Cybersecurity
Buyer-intent questions where a wrong shortlist is expensive.
Semiconductors
Where a chip crosses a border, and where the money lands.
Predictions
Every call we registered, including the ones we got wrong.
Most recent
All three cable networks won election night, and none of those numbers settles a deal
The question: Three networks published a “most-watched” claim from the same Tuesday, and the cable order changed twice in three days. Which one is right, and is any of them the number an advertiser actually buys on?
What it found: All three are right, because each leads a different cut — Fox News on primetime total viewers, CNN on adults 25–54, MS NOW on the full 6pm–2am window. The order changed because three Nielsen releases came out across three days and 45,000 viewers covered all three networks in the first one. And the measure that settles a national television deal is the average commercial minute, usually adults 18–49, which was in none of the releases.
The free window has four shows in it, and CBS Studios made all four
The question: CBS posts full episodes of some fall shows to YouTube for free, two days after they air, while Paramount+ charges for the same episodes a day earlier. What decides which shows get that treatment?
What it found: Two conditions, and together they fit every title examined: CBS Studios produces it, and it is launching. Not one outside-produced title is in the window — eight for eight — which means the network’s most-watched series cannot be put there by deciding to. But the two conditions never disagree on this slate, because every first-season title on it is also a CBS Studios title, so nothing here can say which one is doing the work.
Kraft Heinz is not in the answer to its own category’s questions
The question: A content budget is spent on the assumption that publishing good answers to category questions reaches the household asking. Does it?
What it found: Across nine answers to the questions the US meals category is actually asked, the brand held none. On the largest assistant nothing it publishes was read at all — 318 to 321 input tokens across 2,404,500 monthly searches. Where an assistant did read, on a question about the brand’s own two products, it cited a home-decor site and a salad guide ahead of the brand’s pages. Perplexity cited the brand’s own domain first, so the lever exists — on the operators almost nobody uses.
Google’s AI Overview does not cite the pages Google ranks
The question: Is answer-engine work a discipline of its own, or a second name for search ranking — does the generative answer cite the pages that already rank beneath it?
What it found: It does not. Across 8 high-demand questions, an AI Overview appeared on every one, and its citations overlapped the organic results on the same page by an average of 35%. Two-thirds of what it cites does not rank beneath it. The split runs by question shape: a named dish reached 83%, while a shopping task shared no domain at all.
Most answers are not researched. They are remembered.
The question: Before a content budget is justified on AI visibility, does an assistant actually read anything when answering the questions a brand cares about?
What it found: Not usually. Retrieval leaves a measurable signature — about 8,174 input tokens when pages are fetched against roughly 320 when they are not — and across eight questions carrying 2,404,500 monthly searches the largest assistant left none of it. The rule is not fact against opinion: it tracks whether the answer moves. Prices move, recalls move, recipes do not. Two earlier rules were written, tested and discarded.
Every study lives under its vertical above. Nothing here is a summary — each page carries its own rows.
What makes it a standing study
- The question and its falsifier are written down before any data is pulled. A prediction recorded afterwards is worthless, and a question that cannot come out wrong is not a question.
- Every row is published, not summarized. A citation list reported as a percentage cannot be checked by the reader, and an unverifiable figure is an assertion.
- Every figure carries its source, its method and its date. Including the model version, the location a search was run from, and what the pull cost.
- The limits are on the page, not in a footnote. Including what the instrument can never see.
- It is re-run on a stated schedule and corrected in place, dated. A measurement that can be taken again is taken again.
Everything else on this site descends from these pages. A finding on the blog names the study it came from; a Decision Brief applies one of these instruments to a named company's own decision. The research is free to read and to cite, with no form.
What a Decision Brief does with this
These studies are the instruments. A Decision Brief points one of them at a single seat's six decisions and answers each from evidence, naming who owns it, who contests it, and what would have to be true for the answer to flip.
See the Decision Briefs$495 standard, $995 custom. Pre-made and delivered on payment. If the decision turns out not to be decidable from evidence, it is refunded.
Every study here is re-run, and the next reading will differ from the last.
Sound Decisions carries each re-run and each correction, with the registered question written before the data is pulled.
Subscribe to Sound Decisions