Automated rule: pause video ads on low hook rate
Meta's rules can't read hook rate. How to pause a video on a weak opener anyway: a floor from your own baseline, an impression minimum, a hold-rate check.
Meta's rules can't read hook rate. How to pause a video on a weak opener anyway: a floor from your own baseline, an impression minimum, a hold-rate check.
Thursday morning. Four new UGC videos went live on Tuesday. The fourth has 3,400 impressions, one purchase, and a 13% hook rate in an account that usually sits in the low twenties. Cost per purchase can't tell you anything yet. The opener already has. You open the rule builder to automate the call next time, and the metric isn't there.
Short answer: An automated rule that pauses video ads on a low hook rate needs three parts: a floor from your account's own median for that placement, at least 3,000 impressions before it fires, and a hold-rate check that spares a slow opener with strong retention. Meta's rules cannot read hook rate, so the ratio comes from elsewhere.
The takeaways
No. Hook rate is a homemade metric (3-second video plays divided by impressions, built as in how to check hook rate), and Meta's rules only read fields Meta exposes. The filter list in Meta's Marketing API evaluation spec includes video_play, and nothing for 3-second plays, ThruPlays or custom metrics. Check your own rule builder before you plan around it, since that interface moves.
That leaves three routes: a saved report sorted by your hook rate column that you check every morning, a small script that computes the ratio and pauses through the API, or a third-party rule tool that calculates video ratios itself. The logic below is the same for all three.
Hook rate is the earliest verdict a video gives you. At a €12 CPM, 3,000 impressions cost €36 and the hook rate is readable. A cost-per-purchase rule with a sensible spend floor (I use three times the target CPA, so €120 at a €40 target) waits more than three times as long for its first verdict.
On a video whose first three seconds lose almost everyone, the extra €84 buys a cost-per-purchase number you could have predicted. The catch: hook rate measures attention to the opener and says nothing about the offer, the page or the price. A hook rule should catch the obvious failures early and leave the close calls to the cost rule.
About 3,000 impressions per video, in my experience, and the arithmetic backs it. Treat each impression as a yes-or-no on "stopped for three seconds". At a true hook rate of 25%, the standard error at 500 impressions is about 1.9 points, so the 95% band runs roughly ±4 points. At 3,000 impressions the band shrinks to about ±1.5 points.
The real noise is somewhat wider, because the same person sees an ad more than once. So at 500 impressions a 21% and a 27% hook rate can be the same video. At 3,000 they mostly can't. Make the impression minimum a hard condition, or the rule fires on whatever had a rough first hour.
Your own baseline. Benchmark pages trade one number for every account, but hook rates depend on niche, creative style and placement, so a borrowed floor pauses half your videos or none.
What I do: take the median hook rate of the last 15 to 20 videos that cleared 3,000 impressions in the same placement, and set the floor at about 70% of it. A Feed median of 24% puts the floor near 17%. That catches the openers that clearly failed and leaves the average ones to the cost rule. Recompute the median monthly, because creative styles drift.
A blended hook rate is a weighted average across placements, and Advantage+ placements change the weights for you. Reels hooks run higher than Feed hooks because Reels opens full-screen with the video already playing.
An illustration with made-up numbers: a video hooks 20% in Feed and 32% in Reels. At a 60/40 delivery split the blended rate is 24.8%. Next week delivery shifts to 85/15, and the blend reads 21.8%. Three points gone with the creative untouched, enough for a rule to pause a video doing exactly what it did last week. Read hook rate per placement, or compare against a baseline built on a similar delivery mix.
Because a slow opener on a strong video is the cheapest fix in the account. Hold rate (15-second views over 3-second views) tells you whether the people who stopped stayed. A 15% hook rate with a hold rate well above your median means the body works and only the opener failed. Pausing it throws away the expensive part, so split the rule. Low hook plus low hold: pause. Low hook plus above-median hold: alert, and recut the first three seconds. That second branch is exactly where an alert beats an automatic pause.
Add one more exemption: if the video already has purchases at or below your target CPA, the rule stays quiet. Sales outrank attention.
A hook rule will sometimes be wrong, and a pause can be undone. A paused ad keeps its ID, comments and history, so a wrong call costs a day of delivery. Paused videos are also your recut pile: a body with a decent hold rate under a failed opener is half a new ad.
In Adscalr, rules fire on 8 metrics, hook rate and hold rate among them, so this is a native condition there and needs no script. Actions are pause or kill only, a kill is reversible for 30 minutes, and every evaluation is logged with the number it read. One honest trade-off: the learning-phase lockout blocks any action on an ad younger than 5 days or under €200 of spend, a hook rule included, so it acts later than the 3,000-impression rule above. The safeguards are listed on the Automation page.
This is the thinking behind Adscalr.
See the product →