What Counts as Good Retention on Short-Form Video?

Retention is the metric that decides whether short-form video gets distributed, and most creators look at the wrong part of it. Average view duration is a summary. The shape of the curve is the diagnosis.

The Three Numbers That Matter

Two-second hold. The percentage of viewers still watching after the opening moment. This is the highest-leverage number on the whole graph, because everything downstream only applies to viewers who stayed. If this is low, nothing else you fix will show up in the results.

Halfway retention. The percentage still watching at the midpoint. This tells you whether the middle of your video earns its length. A video that loses half its audience by the midpoint is usually too long for its idea, not badly made.

Completion and loops. On short-form, finishing is a strong positive signal, and re-watching is stronger still. Videos that loop cleanly — where the ending flows naturally back into the opening — tend to overperform for reasons that have nothing to do with content quality.

Rough Benchmarks, With a Caveat

Published benchmarks vary widely and depend heavily on video length, niche, and audience. Treat any specific percentage as a starting reference rather than a target, and build your own baseline from your own videos.

As a general shape: strong short-form videos hold the large majority of viewers through the first two seconds, and keep a substantial share to the midpoint. Shorter videos naturally show higher completion, which is why comparing a 20-second video's completion rate against a 60-second one tells you very little.

The most useful benchmark is your own median. Take your last twenty videos, find the middle performer, and treat that as the line to beat. A video that beats your own median is doing something worth repeating regardless of how it compares to someone else's channel.

Reading the Curve

A sharp early cliff means the promise was wrong. Either the opening did not signal what the video was about, or it signalled something the viewer did not want.

A gradual slide means the video is fine but slow. There is usually a specific sentence where interest breaks — find it, and cut everything before the point that sentence was setting up.

A mid-video spike means viewers scrubbed back to re-watch something. That is a strong signal: whatever happened there is worth building more videos around.

A flat line at a high level with very few views means the video works and was never distributed. That is a topic-demand problem, not a quality problem, and the fix is choosing subjects people are actually looking for.

Which to Fix First

Always the opening. It is the cheapest thing to change, it affects every other metric, and it has the largest effect size. Rewrite the first line of five videos to make a specific claim rather than an introduction, and compare.

Once the opening holds, work on length. Most underperforming short-form videos are twenty to thirty percent longer than their idea justifies.

Only then consider posting time, hashtags, or thumbnails. Those shift distribution at the margins. Retention shifts it fundamentally.

One Video Is Not Data

Retention on any single video is noisy. Judge changes across at least five videos before concluding anything, and change one variable at a time. Most creators who believe the algorithm changed have actually run one experiment with three simultaneous variables and no control.