In video, most of what decides whether an ad works happens in the first few seconds, and CPA does not tell you where it failed. That is what two metrics paid social teams have calculated by hand for years are for: hook rate and hold rate. In late October 2026 they started appearing as native columns in Ads Manager in some accounts, so it is a good time to define them before everyone reads a different number.
Hook rate: did it stop the scroll?
Hook rate = 3-second video plays ÷ impressions. It measures whether the opening (first frame, first line, on-screen text) gets people to stop. If it is low, the problem is the first seconds; do not touch the ending or the landing page yet.
Hold rate: did they stay?
Two formulas are in use; pick one and keep it:
- ThruPlays ÷ impressions. Simple, but a weak hook drags it down too.
- ThruPlays ÷ 3-second plays. The share of people who stopped and stayed for the body. Better for diagnosis, because it separates hook from content.
A ThruPlay is 15 seconds, or the full video if shorter, so very short videos inflate hold rate. Compare videos of similar length.
Reading them together
| Hook | Hold | What it usually means | What to test |
|---|---|---|---|
| Low | High | Those who stop stay, but few stop | A new opening on the same video |
| High | Low | The opening promises what the video does not deliver | Move the core message earlier |
| Low | Low | The concept does not land with this audience | A new angle, not a new cut |
| High | High | The video holds attention | Check CPA: if bad, the problem is after the video |
Hook and hold are diagnostic metrics, not outcome metrics. A great hook that attracts the wrong person produces enviable attention numbers and a bad CPA. Pick winners on cost per result; use hook and hold to decide where to intervene. More in creative testing in paid social.
What values to watch
There is no universal good hook rate; it varies by industry, placement and audience. Use:
- Your own median, by placement.
- The trend of the same ad. A falling hook rate with the same audience is an early fatigue signal, usually ahead of CPA; see creative fatigue: when to refresh.
- Minimum volume. Wait for a few thousand impressions before comparing.
Native columns: check before you switch
The native columns reported in late October have no confirmed formula or universal availability. Put the native and custom columns side by side on the same ads; if they differ, find out which denominator the native one uses; and if you switch definitions, note it in reporting. The same week, some accounts also showed video retargeting audiences as three tiers (all, engaged, highly engaged viewers) instead of percentages; check the thresholds in your own account.
How to improve each
Hook: test new openings on the same video before producing a new one: a different first frame, an on-screen question, the product or result in the first second, a face talking to camera. See static vs video in paid social.
Hold: deliver on the opening fast, cut brand intros, add captions, and get the core message in before 10 seconds. Creator and customer footage often holds better; see UGC vs brand-produced creative.
I followed the platform changes in this note in Yoann Ferrand’s Paid Media News newsletter.
