Average retention on Reels: what it is and why 100% is not the target

In practice, the calculation is average watch time per play divided by video length. A 30-second Reel watched for an average of 12 seconds has 40% average retention.

Being a percentage, it solves the problem total time does not: it lets you compare pieces of different sizes. But it only solves it partly, because retention itself carries a length bias, and that is where most reports go wrong.

One quirk that confuses people: on very short videos retention can exceed 100%. That happens when the loop runs more than once for the same person and watch time exceeds the length of the piece. It is not an error, it is replays.

The longer the video, the more chances to leave. That is attention arithmetic, not a quality judgement.

The practical consequence is that retention is only comparable within a duration band. An 8-second Reel at 85% and a 90-second one at 35% can both be excellent at what they set out to do: the second delivered 31 seconds per person, the first delivered 6.8.

That is why the correct reading is always double: percentage retention to know whether it held, and watch time to know how much attention was actually delivered. Optimising only the percentage pushes everyone towards ever shorter videos, which is not always what the content needs.

There is a case that shows up constantly and almost nobody interprets: extremely high retention with low reach.

It usually means the video was watched in full by people who already follow and already like you, and was not delivered beyond that. It is strong relationship content and weak discovery content. Retention does not expose the problem because it only looks at whoever arrived.

The inverse holds too: mediocre retention with very high reach can be a viral that attracted the wrong audience. In that case the percentage looks bad and the result, in new audience, may have been excellent.

Retention is a content metric. Reach is a distribution metric. Together they tell the story.

• Cut the beginning. The most expensive loss happens in the first seconds, and most of it is intro, greeting and unnecessary context.
• Cut the middle. Retention drops where pacing drops. If you do not know where, the retention curve shows you.
• Match length to the idea. A stretched video loses retention; a video cut to the bone preserves it.
• Work the loop. On short pieces, an ending that flows into the opening adds replays and lifts the percentage.
• Test the hook alone. Changing two things at once turns the test into an opinion.

• I compare retention only between videos of similar length.
• I read it alongside watch time and reach.
• I do not treat 100% as a goal: it can mean the video was too short.
• I use the curve when the average is bad, to find where I lost people.
• I give it a few days before judging, because Reels accumulate.