The first three seconds are the same problem on both platforms, and the data is clearer on one of them
Our corpus is Instagram Reels, not TikTok, and we say so up front. What transfers between them, what does not, and the openings that actually performed. 2026-10-11.
A disclosure before any numbers
Every figure here comes from 2,512 Instagram Reels. We have not analysed TikTok, and we are not going to present Reels data as TikTok data without saying so.
What makes it worth reading anyway is that the hook problem is the same shape on both: a vertical video, autoplaying in a feed, where the viewer decides in about a second whether to stay. The formats, lengths and openings are close enough that the mechanics transfer. What does not transfer is anything about distribution, the For You page, or audio trends, so this page does not make claims about those.
What a hook is competing with
The baseline is harsher than most advice admits. 1,642 of 2,512 Reels, 65 percent, reached fewer people than the account already had followers. A hook is not the difference between good and great; it is usually the difference between being seen and not.
The length edge, which applies on both platforms
The most reliable finding in our sample is not about words at all. Reels under 15 seconds are the worst performing length band we measure. Short videos give a hook nothing to pay off, and the shortest clips are where silent content clusters: the median reel with no speech runs 15.3 seconds.
Four openings that are not a line of dialogue
| Opening | How it works |
|---|---|
| Show the setup, hide the outcome | The frame raises a question the video answers |
| Start mid-action | No preamble, so there is nothing to scroll past |
| Put the contradiction on screen | Two things that should not be together, before a word |
| Name something they have felt but never labelled | A new word is a reason to stay and find out what it means |
These matter because most hook advice is about sentences, and the strongest opening family in our data is the one where the picture carries it. You cannot write your way to that; you have to frame it.
What to do
- Design the first frame before the first line, on either platform.
- Clear 15 seconds. On our data that is the single largest step in performance.
- Stop opening on a count of things. It reads like a hook and performs like the weakest family we measure.
- Measure in views per follower, not raw views, or you are just measuring how big you already are.
- Treat anything we say about distribution as not transferring. The craft transfers; the algorithm does not.
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.
See the openings that actually outperformed, measured against the creator's own audience.
Analyze a creator freePut your own idea through the same checks before you film it.
Create content