Streaming · Answer

Does a big opening mean a streaming title is working?

A big opening tells you about the opening, and almost nothing about what follows. The common measure — demand three months later as a share of the opening peak — is partly arithmetic: a title that spikes 89× above its own baseline has further to fall than one that spikes 11×. Across 28 titles measured, that retention figure correlates with the size of the spike at r = −0.53. Change the denominator and the answer changes. A measure that nets the baseline out of both ends is less confounded (r = +0.30) and reports no difference between the two cohorts we tested: 1.2% against 1.4%, where the retention form reported 12.2% against 4.9%. Two defensible measures, opposite answers. If you are comparing decay across years or across panels, the denominator has to be declared, because it can decide the result.

The evidence

Retention is confounded by spike size.

Correlation of log spike-above-baseline with retention at three months: r = −0.53 across 28 titles. Pooled, titles spiking under 10× retain a median 18.2%; titles spiking 10× or more retain 7.0%.

A baseline-netted measure is less confounded and gives a different answer.

Residual lift — (value at +3 − baseline) ÷ (peak − baseline) — correlates with spike at r = +0.30 and shows 1.2% against 1.4% for the two cohorts the retention form separates.

A pre-existing-volume screen fixes this by construction, and it is already in use.

The published demand-decay study rejects any term carrying 1,000 or more monthly searches before the title existed. Inside that panel, peak size and retention correlate at r = +0.02 — no relationship. The screen was written to stop contaminated terms and also removed this confound.

What would have to be true for this to be wrong

If every title in your panel launched in the same window with no prior season and no pre-existing term, the two measures converge and retention is fine to use. The confound appears when a panel mixes bare franchise terms with season-qualified ones, or compares one year against another.

ADMADS Engine, four runs pulled October 27, 2026, plus the thirteen-title panel published at cultmath.com/research/streaming/demand-decay. Google Ads monthly search volume via DataForSEO Labs, United States, English, from 2019. Baseline is the median of the six months before each title’s peak.
Every row behind this answer is published in the standing study at cultmath.com/research/streaming/netflix-engagement, which is re-run on a schedule and corrected in place, dated.
Published 2026-10-27 by Ipalibo Da-Wariboko, CULT+MATH LLC. Where a number could not be sourced, this page says so and gives none.

This is one of six decisions covered in the streaming decision brief — each with the evidence behind it, who owns it, who it is contested with, and what would have to be true for the answer to flip.

See the brief →