The retention graph in YouTube Studio tells you where viewers left. It does not tell you why, and by the time you can see it the video is published. The Audience Retention Heatmap predicts the curve from your video's structure before you publish, marks the five riskiest drop-off points, writes the script fix for each, and estimates how much retention you recover by applying them.
This guide explains how the heatmap is built, the five recurring danger zones, and how to use it with the real graph after publishing.
How the heatmap is built
You describe the video as a sequence — intro, sections, reveals, sponsor, CTA, outro — with rough durations. The model overlays that on the retention shapes typical for your format and duration band and highlights where your structure creates a natural exit: a long setup, a section that explains before it shows, a mid-video ad read, an early CTA, a payoff that arrives too late. Each point gets a heat score from the expected drop and the number of viewers still present to lose.
The five danger zones
Zone one is the first 30 seconds — greeting, logo, over-long context. Zone two is the setup-to-payoff gap, usually between 20% and 40% of the runtime. Zone three is the mid-video sponsor or tangent. Zone four is a repeated or redundant section. Zone five is the outro — a CTA that signals "the content is over" before it is. The rewrites are concrete: cut, reorder, tease, move, or merge.
After publishing
At 48 hours, put the real Studio curve next to the predicted one. Where they agree, you have confirmed the structural weakness and can fix it in the next video of that format. Where the real dip is somewhere the model did not expect, it is a delivery or editing issue rather than a structural one — check the footage at that timestamp.
Tips
- Include timings even if rough; a heatmap without durations is just a list of sections.
- Say where the single most surprising moment is — the strongest fix is usually to move or tease it.
- Run it on your last three videos of the same format to see the pattern you repeat.
Common mistakes
- Fixing the outro first — Almost nobody is left by then. Fix the earliest hot zone.
- Adding a mid-roll to a 9-minute video — The revenue is small; the retention cost is not.
- Treating the prediction as analytics — It is a structural risk map. The real graph is the truth.
Worked example
Input: "13-min budget grinder review: intro 70s, unboxing 2.5m, grind tests 5m, sponsor 60s at 7:00, taste test 2m, verdict + CTA 1.5m."
What the report returned: Hot zones: 0:00–1:10 (cut intro to 20s), 3:40–6:00 tests before any result (state first result at 4:00), sponsor at 7:00 (move to 9:30), verdict at 11:30 (state winner first). Est. AVD 44% → 55%.
Frequently asked questions
Does it use my actual retention data?
No — it predicts from structure so you can fix videos before publishing. Compare it with Studio's real graph afterwards.
What is a good average view duration?
Above 50% for videos under 10 minutes, above 40% for 10–20 minutes, in most niches.
Can it analyse a published video?
Yes, by URL. It uses the public duration and chapters as the structure.
How is it different from the Watch Time Optimizer?
Watch Time Optimizer focuses on length and three top edits. The Heatmap maps five zones with scored severity and is more granular for longer videos.