Banking baseline before 24 August
COLLECTING
From 2026-08-24 YouTube counts a view from the first frame of playback. Earnings and Partner Programme eligibility stay tied to engaged views, which are not shown publicly. Every rate card, CPM and performance clause written against “views” now means something different from what it did last week.
A cumulative view total spans a video’s whole life, so comparing totals across 24 August would measure nothing. Plotted here is accrual: views gained per video per hour, differenced from hourly snapshots of one fixed set of videos. A counting change shows up as a step.
If you priced a creator deal against public views, this is the gap between what you bought and what you paid for.
The engaged-view rows unlock once both sides of the 2026-08-24 change have been observed. Nothing here is modelled or estimated in the meantime — the multiple is measured or it is blank.
We re-price live YouTube creator deals against engaged views, and give you the clause language to stop it happening at the next renewal. Fixed fee, one week, no platform to buy. For agencies and brands running YouTube creator spend.
| Category | Channels | Subs | Pre | Post | Inflation |
|---|---|---|---|---|---|
| beauty | 25 | 82M | 28.37 | — | — |
| education | 25 | 168M | 61.94 | — | — |
| entertainment | 26 | 696M | 2,080.82 | — | — |
| finance | 25 | 26M | 3.92 | — | — |
| fitness | 25 | 109M | 39.03 | — | — |
| food | 25 | 109M | 30.04 | — | — |
| gaming | 25 | 423M | 254.77 | — | — |
| music | 25 | 581M | 154.07 | — | — |
| news | 26 | 145M | 5.67 | — | — |
| science | 25 | 146M | 213.62 | — | — |
| sports | 25 | 258M | 0.11 | — | — |
| tech | 25 | 129M | 70.59 | — | — |
Pre and Post are views gained per video-hour, mature cohort.
Panel. 302 channels across 12 categories, 302 registered before the change and therefore eligible for the headline. Seeds skew large, so the panel is deliberately topped up with 50k–5M subscriber channels. Counters are read hourly.
Cohorts. The headline uses videos published on or before 2026-07-25 — 30+ days old at the change, where view velocity decays smoothly. Newer uploads are tracked but their decay is steep enough to swamp the effect, so they are reported separately and never in the headline.
Adjustment. Views decay roughly as a power law in video age. The decay curve is fitted on the pre-change window only, extrapolated across the change, and the headline is the ratio of observed to predicted accrual. The interval is a percentile bootstrap resampling videos, not intervals, because intervals from one video are correlated.
The limitation that matters. There is no control group. The change is global and simultaneous, so nothing was left untreated to compare against. This is a discontinuity estimate resting on an extrapolated decay curve — not a randomised comparison. If YouTube shipped anything else the same week, or if the rollout is staged by region rather than instant, that lands in this number and cannot be separated out from public data alone.
41 downward revisions were observed and excluded from accrual: YouTube purges invalid views retroactively, so counters do go backwards. Their frequency is reported rather than hidden.
Full methodology, the channel panel and the raw series are available on request — jesse@engagedviews.com.