Instagram does not show you anyone else's retention, and most tools that claim to are estimating
Our own retention figure takes only 26 distinct values across 2,512 Reels, and three of them cover 77 percent of the corpus. An honest account of what retention data exists. 2026-10-11.
What retention actually means here
Retention is the share of a video people stay for. On your own account Instagram shows you a version of this in Insights, as average watch time and a retention graph. That is real measured data and it is the thing worth looking at.
On an account you do not own, it does not exist. Instagram publishes views, likes, comments and follower counts. It does not publish completion rate to third parties. Any tool showing you a competitor's retention is modelling it.
Including ours, and here is the evidence
Scrptly reports a retention score. It is an AI estimate on a 50 to 95 scale, not measured data, and the shape of the numbers gives it away. Across 2,512 Reels the score takes only 26 distinct values, and three of them account for 77 percent of everything.
Measured completion data would be continuous. A handful of repeated values is what a model produces when asked for a number in a range.
| Score | Reels | Share of corpus |
|---|---|---|
| 82 | 811 | 32.3% |
| 78 | 713 | 28.4% |
| 88 | 407 | 16.2% |
| 72 | 185 | 7.4% |
| 75 | 130 | 5.2% |
The part most tools would not tell you
Those scores do correlate with performance. Reels scored 88 have a far higher median view count than Reels scored 78. It would be easy, and flattering, to present that as proof the estimate predicts reach.
It is not proof, because the correlation is circular. We checked the prompt: the analyser is given the view count, like count and comment count before it is asked to estimate retention. A model told a video has two million views will not call it low-retention. The score is partly a restatement of the number it was shown.
So the honest position is that our retention score is useful for sorting your own drafts against each other, and is not evidence that any particular video held an audience.
What you can measure instead
| Signal | Available for other accounts? |
|---|---|
| Views, likes, comments, follower count | Yes |
| Length, caption, audio, posting time | Yes |
| Completion and retention rate | No |
| Average watch time | No |
| Shares and saves | No |
| Your own retention graph | Yes, in your own Insights only |
The proxy that actually works
With what is visible you can build views per follower, which is the only measure that survives comparison between accounts of different sizes. In our sample 490 of 2,512 Reels beat three views per follower, and 1,642 reached fewer people than the account had followers.
That is not retention. It is reach relative to audience, and it is real rather than modelled, which makes it the better number to plan against.
What to do
- Use your own Insights for retention. It is the only measured retention you will ever have.
- Treat every third-party retention figure, ours included, as an estimate until the tool says where it came from.
- Ask any tool quoting retention whether its model sees the view count. If it does, the number is partly circular.
- For competitors, compare views per follower instead. It answers the question you were really asking.
Figures on this page come from 2,512 Instagram Reels across 530 accounts, analysed by Scrptly and deduplicated so no reel is counted twice. Read 2026-10-11.
Compare any creator's Reels in views per follower, which is measured rather than modelled.
Analyze a creator freePut your own idea through the same checks before you film it.
Create content