Hook rate - the share of impressions that survive three seconds

Hook rate is the share of a video ad's impressions that turn into 3-second video plays: 3-second plays divided by impressions. It isolates one thing (whether the opening frames stopped the scroll) and it is the first number a new creative earns. Hold rate is then measured only on the people the hook kept, and click-through rate and cost per acquisition are earned almost entirely from them even though both are still reported against every impression. That is why a weak hook rate is a verdict on the opening, not on the offer. Media buyers use it as the kill signal in creative testing: change the first seconds, not the landing page.

Formula

Hook rate = 3-second video plays ÷ impressions × 100. Thumbstop ratio is the same calculation under a different name, so the two numbers are interchangeable as long as both sides of the comparison use the same denominator. The trap is the denominator: reach counts people, impressions count deliveries, and dividing plays by reach inflates the rate on any ad with meaningful frequency. The 3-second threshold is Meta's: a 3-second video play is counted at three seconds, or at 97% of the video when it runs shorter than that. TikTok's video play metrics list 2-second and 6-second video views and no 3-second column, so a TikTok hook rate is a different number against a different threshold, comparable to itself over time, not to a Meta hook rate.

Worked Example

An ad delivers 84,000 impressions and records 12,600 three-second plays. 12,600 ÷ 84,000 = 0.15, a 15% hook rate. A second variant of the same ad with a different opening delivers 61,000 impressions and 6,100 plays: 10%. The two variants share the body, the CTA and the offer, so the five-point gap is the opening and nothing else. At the same CPM the first variant hands the rest of the ad half again as many viewers to work with.

Why We Do Not Publish a Hook-Rate Benchmark

A benchmark is only worth anything if the person publishing it measured it. Admetra analyzes ads from public Meta ad delivery, where impressions and 3-second plays are not disclosed, so we have no corpus hook rate to quote and will not borrow one. What we do measure is the thing a hook rate is a proxy for: where the opening ends, what it does, and how the ad's delivery ranked inside its own brand.

How Admetra Bounds a Hook

Every analyzed ad is cut into beats with exact timestamps, and the hook is the only beat with a constraint on where it may start. A hook that opens late, or runs long enough to have become the body, is treated as a segmentation defect rather than a stylistic choice, because a hook that cannot be lifted out and reused is not a hook. One vocabulary note, because we take a position on it elsewhere: most teams use "hook" for whatever stops the scroll, and hook rate is named for exactly that. Our scroll stopper entry separates the interrupt that wins the stop from the promise that keeps someone watching. That split is our own editorial convention rather than standard usage, and it does not change what this metric counts.