If you are reviewing your own content, score retention before anything else. It separates further than the hook does
Retention is the strongest single predictor we have measured. 0.32 views per follower at 70 to 79, 2.08 at 90 plus. Measured 26 September 2026.
What the numbers say
A content review is only worth doing if you know which column to read first, and after five weeks of measuring this corpus the answer is retention rather than the hook. Across 2,335 Reels banded by their retention score, the 70 to 79 band reached a median of 0.32 times their follower count and the 90 or more band reached 2.08. That is about 6.5 times, and it is a wider separation than any other single measure we have found. The hook score, which is the thing most advice points at, separates 0.35 from 1.37 on the same corpus, about 3.9 times. Both matter. Retention matters more, and the reason is not mysterious: a strong opening buys you three seconds, and retention is whether you did anything with them.
When you review a video, do not stop at whether the first line was good. Score whether the middle earned the end. That is the part with the 6.5 times spread on it, and it is the part a review usually skips because it is harder to judge.
Sample of 2,335, measured on 2026-09-26. Scrptly production database, queried 2026-09-26. 2,335 Reels deduplicated by shortcode, each carrying a retention score, a view count and its account's follower count. IMPORTANT: this corpus is Instagram Reels, not TikTok. No TikTok video was measured.
The measurement
Retention score is Scrptly's own 0 to 100 score for how well a video holds attention across its runtime, assigned at analysis time before any reach figure is looked at. Views per follower is the video's views divided by its own account's follower count. The five bands hold all 2,335: 66 plus 978 plus 778 plus 485 plus 28. Two cautions. The bottom two bands are effectively tied, so the useful signal starts at 80. And the top band rests on 28 videos, which is why its top tenth column reads lower than the band below it: that is small-sample noise, not a ceiling.
| Retention score | Reels measured | Median views per follower | Top tenth reaches |
|---|---|---|---|
| Under 70 | 66 | 0.35 | 5.57 |
| 70 to 79 | 978 | 0.32 | 5.17 |
| 80 to 84 | 778 | 0.57 | 11.00 |
| 85 to 89 | 485 | 0.96 | 14.55 |
| 90 or more | 28 | 2.08 | 7.62 |
One Reel, beat by beat
What an 88 looks like when you take it apart. One real video from the corpus, not a composite: 9,127 followers, seen 2,477,185 times, which is 271.41 times the following. It runs 74 seconds, which is long, and it holds because every beat changes the kind of information being given rather than continuing the last one. Review your own videos against this shape: shock, then turn, then proof, then hand it back.
| At | What happens |
|---|---|
| 0:00 to 0:04 | Ask the large hypothetical straight out. No setup, no throat clearing |
| 0:04 to 0:15 | Show the immediate chaos. This is the answer the viewer expected, delivered fast so it can be overturned |
| 0:15 to 0:35 | The turn: the unexpected positive outcomes, with figures attached |
| 0:35 to 1:00 | Stack specific examples. This is the longest stretch and it is doing the credibility work |
| 1:00 to 1:14 | Hand it back as a question the viewer has to answer for themselves |
What to do next
- Review in this order: did it hold, then did it open well, then everything else. The hold is where the wider spread is and it is the harder question, which is why it gets skipped.
- Judge holding by beats rather than by feel. Write out what changes at each point in the video. If two consecutive beats are doing the same job, that is where people left.
- Be honest about the sample when you use these numbers. The 90 or more band is 28 videos. The bands that carry real weight are 80 to 84 and 85 to 89, at 778 and 485 videos, and the step between them is 0.57 to 0.96.
- One limit stated plainly, since this page is about TikTok and the corpus is not: every figure here comes from Instagram Reels. The structural lesson carries across because the constraint is a vertical feed and a scroll, but no TikTok video was measured, and anything TikTok specific such as its search or its sounds is outside what this page can tell you.
- Scrptly scores retention and the hook separately on every video it analyses and returns the beat order, which is what makes a review repeatable rather than a matter of taste. The first analysis does not need an account.