Two numbers do most of the work when you judge a video ad before the purchase data is in: hook rate and hold rate. One tells you whether people stopped. The other tells you whether they stayed. Read on their own, each one lies a little. Read together, they tell you exactly which part of the video to fix.
The two formulas
Meta does not report either number directly, so you build them from columns it does report.
A 3-second video play is counted when someone watches at least three seconds (or the whole video if it is shorter). A ThruPlay is counted when someone watches to the end, or at least 15 seconds of a longer video.
Some teams define hold rate as ThruPlays divided by impressions instead. That version mixes the hook back in: a weak opening drags it down even when the middle of the video is great. Dividing by 3-second plays isolates the part of the video that comes after the hook, which is the point of having two numbers. Whichever you pick, pick one and use it everywhere, or your comparisons are meaningless.
Paste impressions, 3-second plays and ThruPlays for a few ads and compare them side by side.
What "good" looks like (and why to be careful)
You will see rules of thumb like these quoted a lot:
| Weak | Okay | Strong | |
|---|---|---|---|
| Hook rate | under 20% | 20% to 30% | 30% and up |
| Hold rate | under 25% | 25% to 40% | 40% and up |
Treat them as a starting point, not a verdict. Both numbers move with things that have nothing to do with the creative:
- Placement. Reels and Stories autoplay full screen and behave differently from Feed.
- Video length. A 9-second video reaches "ThruPlay" at the end; a 60-second video needs 15 seconds of attention. Long videos will show lower hold rates by construction.
- Audience temperature. Retargeting audiences already know you and stop more often.
- Format. Static images have no hook or hold rate at all, so never average them in.
The most useful benchmark is your own account: the median hook and hold rate of your last 30 to 90 days of video ads, split by placement if you can.
Read them together: the four quadrants
The value is in the combination. Put hook rate on one axis and hold rate on the other and every video falls into one of four boxes.
High hook, high hold. The opening earns attention and the body keeps it. If it is also converting, this is a candidate to scale and to iterate on (new hooks on the same body, same hook on new products).
High hook, low hold. People stop, then leave. The first three seconds promised something the rest of the video does not deliver, or the pacing drops right after the opening. Keep the hook, rework seconds 3 to 15: get to the product faster, cut the slow setup, and make sure the story pays off what the opening implied.
Low hook, high hold. The few people who stop love it. The body is good and the opening is hiding it. This is the cheapest fix in advertising: test three or four new openings on the same video. Lead with the strongest visual from later in the cut, put the key claim in on-screen text in the first frame, or open on a face mid-sentence.
Low hook, low hold. Do not polish this one. Check whether the concept or the audience is wrong before spending more editing time on it.
When you iterate, change either the hook or the body, not both. If you change both and the numbers move, you will not know why.
Where these numbers mislead you
They are attention metrics, not money metrics. A video can hook and hold beautifully and still sell nothing, usually because the attention is going to the wrong idea (a funny bit that has nothing to do with the product). Always check spend, purchases and ROAS before calling a winner. Hook and hold explain why an ad is or is not working; they do not tell you whether it is.
Small numbers swing hard. An ad with 2,000 impressions can show a 35% hook rate by luck. Wait for a few thousand impressions per ad before you compare, and compare ads that ran in the same period to the same audience where possible.
Averages hide the story. An ad set average of 24% can be one 40% video carrying three 15% ones. Look at the ads individually.
Track them over time, not once
The first week of an ad is rarely the full picture. Hook rate tends to drift down as the same people see the same opening again, which is often the earliest sign of creative fatigue, before CTR and ROAS slip. If you can, store the daily 3-second play and ThruPlay counts and recompute the rates over any window you like, rather than saving the rates themselves (averaging percentages across days gives the wrong answer when volume changes day to day).
How Fovea handles this
Fovea syncs your Meta ad account nightly and stores the daily 3-second play and ThruPlay counts, so hook and hold rate are recalculated correctly for the 7, 30 or 90-day window you pick on the Stats page. Link an ad to its creative in the library and the creative's own Performance tab shows its spend, ROAS and engagement next to the file itself. Linking is done by hand in the asset panel today; automatic ad-to-creative matching is Coming soon
The short version: hook rate tells you which openings to keep, hold rate tells you which middles to rewrite, and money tells you which ads to keep paying for.
Written by The Fovea team.