We took 4,385 viral Reels apart frame by frame. "Make it shorter" is half wrong.
Short hooks get 3.1x the views. Reels under ten seconds get a sixth. An original study of 4,385 Instagram Reels analysed frame by frame.
Every guide tells you the same two things. Hook them fast. Keep it short.
We can now check both. Scrptly has analysed 4,385 Instagram Reels frame by frame, pulling out the exact opening line, the pacing and the view count for each. They were posted between March 2022 and November 2026, and the median video in the set has 143,800 views, so these are not small accounts.
One of those two pieces of advice holds up. The other is costing people views.
1. Short hooks win, and it is not close
We measured the opening line of every video by word count.
| Hook length | Videos | Median views | Share passing 1M |
|---|---|---|---|
| 1 to 4 words | 348 | 342,814 | 32.8% |
| 5 to 8 words | 1,099 | 140,437 | 26.1% |
| 9 to 12 words | 1,377 | 162,507 | 24.9% |
| 13 to 18 words | 1,145 | 135,935 | 18.9% |
| 19 or more words | 396 | 110,350 | 18.2% |
A hook of four words or fewer carries 3.1 times the median views of a hook of nineteen or more, and is nearly twice as likely to pass a million. The drop is not a cliff, it is a slope. Every step longer costs you.
Four words is less than it sounds. "I was today years old." "Nobody talks about this." "This took eleven months." That is the whole hook.
2. But videos under ten seconds collapse
Here is the half that is wrong.
| Video length | Videos | Median views | Share passing 1M |
|---|---|---|---|
| Under 10 seconds | 328 | 30,836 | 10.7% |
| 10 to 20 seconds | 470 | 203,801 | 27.4% |
| 20 to 30 seconds | 368 | 160,492 | 27.2% |
| 30 to 60 seconds | 1,366 | 189,245 | 27.7% |
| 60 seconds or more | 1,791 | 143,713 | 21.3% |
A Reel under ten seconds gets about a sixth of the median views of a ten to twenty second one. Only one in ten passes a million, against roughly one in four for everything between ten and sixty seconds.
And look at the bottom row. Videos over a minute still hold 143,713 median views. Longer is not the problem. Too short is.
The plausible reason is watch time. A seven second video cannot hold anybody long enough to signal that it is worth showing to more people, however good it is. There is no room to earn the distribution.
So: short hook, not short video. Those are different instructions, and they have been collapsed into one piece of advice that only half works.
3. Numbers in the hook do worse, not better
The listicle hook has been standard advice for a decade. Hooks containing a digit averaged 106,208 median views across 425 videos. Hooks without one reached 150,236. That is 29 percent fewer views, and we expected the opposite.
Our best guess is that a number front-loads the structure. "5 ways to..." tells the viewer exactly how long this will take and exactly what shape it has, which is an invitation to decide they already know enough.
4. Two pieces of standard advice do nothing at all
This is the part we found most useful, because it frees up attention.
| Advice | With | Without | Verdict |
|---|---|---|---|
| Say "you" | 143,713 | 143,890 | No effect whatsoever |
| Say "I" | 136,879 | 152,812 | Slightly worse |
| Ask a question | 170,447 | 141,286 | Plus 21 percent, the only one that earns it |
"Speak directly to the viewer" is in every guide ever written. Across 1,365 videos that use it and roughly 3,000 that do not, the gap is 177 views on a median of 143,800. That is nothing.
What this does not say
- This is correlation, not causation. Short hooks do not cause views. It is entirely possible that people who write tight hooks are better at everything else, and the hook is a symptom rather than a cause. We cannot separate those with this data and we are not going to pretend otherwise.
- The effect is not universal. The short-hook advantage holds in most categories but reverses in two of them, social media content and fitness. If you make that kind of video, treat the headline finding as a prompt to test rather than a rule.
- We did not measure the thing everyone wants measured. We have view counts, not retention curves. We cannot tell you the second people leave.
Method
- 4,385 Instagram Reels, analysed frame by frame by Scrptly between 2022 and 2026.
- 4,365 had both an opening line and a view count, which is the set every figure above uses.
- Posted 31 March 2022 to 3 November 2026.
- Median, never mean. View counts are wildly skewed, and a single 150 million view outlier would move an average and tell you nothing.
- Hook length is the word count of the transcribed opening line, not a judgement.
- Every property measured here is objective: a word count, a digit, a question mark, a duration. We deliberately did not use our own model's quality scores, because a study that grades its own homework is not a study.
Scrptly takes any Reel apart frame by frame: the hook, the pacing, the structure underneath. The same analysis behind this study, run on any video you point it at.
Analyse a Reel free