Guide
Retention rate is the share of viewers still there at each second
Retention rate is the share of viewers still watching at a given point in a video, usually read as a curve from first second to last. Platforms also report it collapsed into one number, average percentage watched. The curve is a map of exactly where people left, which makes it the most actionable chart in video.
- Forms it takes: a curve and an average
- 2
- Curve shapes that each name a different problem
- 3
- Of your own videos make a usable baseline
- 10
One number summarises, the curve diagnoses
The calculation at any second is viewers still watching divided by viewers who started. String those points together and you have the retention curve.
The collapsed version, average percentage watched, is what dashboards surface first. It ranks videos against each other and hides everything about why.
The curve keeps the why. Every slope is people leaving at a rate, and every cliff is a timestamp where something specific lost them.
That specificity is the value. Most creative metrics say how much. The retention curve says where, and where is the only thing an editor can act on.
Three shapes, three different problems
Retention curves fail in recognisable shapes, and each shape names its own repair.
- The opening cliff. A steep drop in the first seconds. The hook failed, or the ad reached people it should not have. Nothing later in the video matters until this is fixed.
- The steady slide. No cliffs, just constant leakage. A pacing problem. The video is losing a race against the scroll every single second, usually for want of change on screen.
- The mid-video cliff. A sharp drop at one timestamp. Something specific happened there. A boring stretch, a pivot to selling, a claim that broke trust. Watch that second and it names itself.
Reading the curve like an editor
Each shape points at a different part of the work. The mid-video cliff is the most repairable of the three.
Opening cliff
The hook failed
Steady slide
Pace too slow for the feed
Mid-video cliff
One moment lost them
Flat tail
Survivors watch to the end
This is the whole editor
Highlight a phrase and a clip lands on those exact words. No timeline, no keyframes, no layers.
The traps are comparisons the number does not support
Length distorts everything. A fifteen second video retains a higher share than a two minute one by arithmetic alone, so retention only compares videos of similar length.
Loops inflate the top. Where short videos replay automatically, watch time past 100 percent flows into the figures, and platform reporting varies on how. Treat cross-platform comparisons as meaningless.
Averages hide cliffs. Two videos can share an average watched figure where one bleeds steadily and the other loses half its audience at one bad moment. Only the curve tells them apart.
And retention is not results. It measures staying, not persuasion. An ad can retain beautifully and sell nothing. Read it beside conversion measures, as the diagnostic layer rather than the verdict.
Improving it is editing, and the levers are known
The repairs map straight onto craft. An opening cliff means a new hook, which is a rewrite rather than an edit. A steady slide means pace: tighter trims, more image changes, captions moving word by word.
A mid-video cliff means watching the offending second and cutting or reworking what happens there. Usually it is a stretch where nothing changed or the selling started too early.
In Cutroom pace is a lever rather than a re-edit. Change it and the machine re-cuts the whole video, so testing a faster cut of the same take against the original costs an export rather than an evening.
Build a baseline before optimising anything. Your own last ten videos, similar lengths, same placement. A curve is only high or low against your own median, never against a number from a blog post.
Questions people ask
- What is a good retention rate?
- There is no universal figure worth trusting, because length, placement and loop counting all move the number. The usable benchmark is your own median across recent videos of similar length. Beating it means the change you made worked.
- How is retention different from hook rate?
- Hook rate reads the opening seconds alone, as a share of impressions that pass an early threshold. Retention describes the entire watch. A video can hook brilliantly and bleed out immediately, and the pair together tells you which problem you have.
- How is it different from watch time?
- Watch time is total minutes accumulated, which scales with reach and length. Retention is a share, which isolates how well the video held whoever it reached. A viral flop can have huge watch time and terrible retention.
- Does retention affect distribution?
- Platforms reward videos people keep watching, so better retention generally buys cheaper reach, organic and paid. The causality runs through viewers, which is why gaming the metric with loops and tricks decays so fast.
The curve is a map of every second you lost someone, which makes it instructions. Cutroom turns the usual fixes, tighter trims and a faster pace, into levers on a cut that already exists.