How hold rate is calculated
Divide ThruPlays by three-second video plays. A ThruPlay is counted when someone watches a video to the end, or for at least fifteen seconds if the video is longer than that. So hold rate takes only the people who did not scroll straight past and asks how many of them stayed with it.
Like hook rate, this is a column you build rather than one Meta hands you. Set it up as a custom metric in Ads Manager so the formula is fixed and everyone reading the report is using the same one. The denominator is the important choice: using impressions instead of three-second plays turns it into a completely different measurement.
Why hold rate matters
It isolates the middle of the video, which is the part that carries the argument. Hook rate tells you whether the opening stopped anyone; hold rate tells you whether what followed was worth stopping for. Between them they turn a vague complaint about creative into a specific brief for the editor.
It also protects you from a false win. An opening built purely to arrest attention will lift hook rate on its own, and the damage only shows up here, as people who stopped for the first frames leave as soon as they see what the video is actually about.
Where hold rate goes wrong
The first problem is comparing it across video lengths. The fifteen-second rule means a short video reaches a ThruPlay by being finished, while a long one has to survive a fixed stretch of time. Those are different tests, and the shorter video will usually look stronger for reasons that have nothing to do with the creative.
The second is reading it without volume. When only a handful of people got past three seconds, the ratio swings wildly and tells you nothing. Let the ad accumulate delivery before you act on the figure.
The third is treating a strong hold rate as proof of a good ad. Plenty of videos hold attention and sell nothing, particularly explainer-style creative that satisfies curiosity instead of prompting an action. Keep cost per result in the same view.
How to act on it
Find the moment people leave rather than guessing. Meta’s video retention chart shows where the audience drops away, and the shape usually points at one cause: a slow section after the opening, a shift in pace or presenter, a long stretch before the offer appears, or a video that has simply said everything it has to say and keeps going.
Then edit for that. Move the strongest proof earlier, cut the setup, keep captions running so a muted viewer can follow, and end shortly after the call to action rather than trailing off. Test the recut against the original and compare hold rate, average watch time and cost per result together — a recut that improves all three is a real improvement, and one that improves only the ratio is not. The wider approach to building and rotating these variations sits with Meta video advertising.