What is hold rate in Meta ads?
What is hold rate in Meta ads?
Hold Rate is an operator ratio: ThruPlays (completion or 15 seconds) divided by 3-second video plays. Meta documents both metrics. There is no official “good” hold rate. Compare variants in the same account and window. If hold dies after a strong hook, fix the body in eonik on Mac; you approve; eonik never spends.
What is hold rate in Meta ads?
Hold Rate is an operator ratio: ThruPlays (completion or 15 seconds) divided by 3-second video plays. Meta documents both metrics. There is no official “good” hold rate. Compare variants in the same account and window. If hold dies after a strong hook, fix the body in eonik on Mac; you approve; eonik never spends.
Hold rate in one line
Hold rate is ThruPlays divided by 3-second video plays. ThruPlay is video completion or at least fifteen seconds of watch time. Or nearly full length if the video is shorter. Meta documents both numerator and denominator.
High hook plus low hold usually means bait-and-switch: the opening promised something the body did not deliver quickly enough. That is a body problem, not another hook swap.
Always pair hold with hook rate
Hook rate alone rewards openings that lie. Hold rate alone hides weak posters. Together they separate stop from stay on the same file. The operator formula for the opening is hook rate.
When you compare variants, lock placement and date range. A hold read on Reels is not the same cell as Feed unless you planned that confound. If frequency is the story, read how ad fatigue actually works.
When hold is the lever
Cut dead air after the hook. Deliver proof before a long brand intro. Match spoken line, on-screen type, and captions so sound-off viewers still get the payoff.
Assemble body fixes in eonik on the same locked hook if the test is body-only. You still watch every export. The generate → assemble loop is how to make AI ads. eonik does not rewrite Ads Manager rules.
Hook vs hold. What each fixes
| Signal | Low reading | Production move |
|---|---|---|
| Hook rate | Opening does not stop scroll | New hook family on locked body |
| Hold rate | Body loses people who stopped | Tighter proof, faster cut |
| Both weak | Whole cell is wrong | New structure, not a tweak |
ThruPlay is completion or ≥15s (≥97% if shorter). Hold rate is operator math. Compare the same account and window.
What usually breaks
- Swapping hooks forever when hold is the problem.
- Using someone else’s hold percentage, not a published Meta benchmark.
- Reading hold on a different placement than hook.
Consider the next cut. Make it in eonik on Mac, approve every frame, then upload yourself. The loop is how to generate AI ads. eonik never spends.
“Copy cannot create desire for a product. It can only take the hopes, dreams, fears, and desires that already exist in the hearts of millions of people, and focus those already existing desires onto a particular product.”
That principle still holds. The next on-brand cut is yours to approve.
Sources
FAQ
What is the difference between hook rate and hold rate?
Hook rate measures the opening three seconds: 3-second video plays divided by impressions. Hold rate measures body retention: ThruPlays divided by 3-second video plays. High hook plus low hold means bait-and-switch. The hook promised something the body did not deliver; Fix the body or stop using that opening family; Meta documents both formulas; neither has a universal good number.
What is a good hold rate for video ads?
Meta does not publish a universal good hold rate. Compute ThruPlays divided by 3-second video plays and compare to your last period in the same account. A drop versus your baseline means the body is losing people who already stopped for the hook. That is operator experience, not a published benchmark; eonik assembles body fixes; you still approve and upload.
How do I improve hold rate on Meta ads?
Deliver on the hook promise in the body, not after a brand intro. Avoid logos and generic product footage between hook and payoff. Cut faster on mobile. Dead air kills hold even when the opening worked; That is operator experience, not a Meta rule; Assemble tighter proof in eonik on Mac; You still decide what launches; Nothing auto-scales spend.
Free, no card · macOS 15+ · Apple silicon