What a retention curve shows
A retention curve is a graph that plots, second by second, the percentage of viewers who are still watching a video. It starts at 100% and falls from there. The shape tells you where attention leaks: a steep cliff in the first seconds points at the opening, while a sudden drop mid-video points at a specific moment worth recutting.
It is the detailed view behind single-number metrics. Watch time and view-through rate compress the whole video into one figure, while the curve shows where the losses happen. YouTube Studio, TikTok analytics and Meta's video reporting each expose a version of it, with different levels of detail for organic posts and ads.
Using it on creator videos
Read the curve against the script. Mark where the hook ends, where the product appears, where the proof lands and where the call to action starts, then look at what the line does at each marker. A drop exactly at the product reveal suggests the reveal feels like an ad; a flat stretch after it suggests the demo is earning attention.
Organic posts from a creator's own account are a useful source, because the creator can share the retention view for videos you are considering licensing. A creator whose curves stay flat through the middle usually knows how to pace content, which is worth more than a high follower count.
Worked example
A 30-second hair-care video keeps 70% of viewers at second 3, slopes gently to 50% by second 10, then drops to 28% at second 12, when the creator starts listing ingredients. The rest of the line is nearly flat.
The ingredient list is the problem, not the hook. The editor cuts it to one line of on-screen text, moves the before and after shot forward and relaunches. The new version keeps 45% at second 12, and the creator is briefed to keep claims visual in the next round.
Common mistakes
Expecting a flat line is unrealistic. Every short-form video loses most viewers early, so judge a curve against your own past videos of similar length rather than an ideal shape. The useful signal is an unusual cliff, not the normal slope.
Another mistake is reading the curve without the script beside it, which turns a precise diagnostic into a vague feeling. Annotate the timestamps, send the creator the exact second that lost people, and plan the next draft with the UGC script template so the fix lands in the brief.
How it's used
"Retention falls off a cliff at second 12, right when she starts listing ingredients; cut that to on-screen text."
"Can you share the retention view for the March video before we license it?"