Hook rate, hold rate, outbound click rate and the two others worth tracking, with 2026 benchmarks and the specific creative change each bad number points at.
HighQualityUGC Team
Editorial
We run UGC ad tests daily and publish what holds up: real credit costs, real hook rates, no vendor fluff.
Cost per acquisition tells you an ad is bad. It never tells you which part is bad, and "make a better ad" is not a brief.
The metrics below are diagnostic. Each maps to a specific section of the video, so a bad number produces an actual instruction instead of a feeling.
| Metric | Definition | 2026 target | What a bad number means |
|---|---|---|---|
| Hook rate | 3-second plays / impressions | 30% or above on Meta | The first two seconds. Rewrite the opening. |
| Hold rate | Completions or 15s views / 3-second plays | Above 45% on a 30s video | The middle. The promise was not kept. |
| Outbound CTR | Clicks that leave the platform / impressions | Above 1% cold, above 2% warm | The ask, or a mismatch with the offer. |
| Frequency | Impressions / people reached | Watch above 3 on a short flight | Not a creative fault. It is fatigue. |
| CPA | Spend / conversions | Account specific | Something upstream. Read the other four first. |
The reason to look at these in order is that each one conditions the next. A hold rate calculated over a tiny number of 3-second plays is noise, and a CPA judged before you know whether anyone watched is meaningless.
The two-by-two that comes out of hook rate and hold rate is the most useful diagnostic in the format:
That last quadrant is worth naming because it is where creative teams waste the most time, redoing a video that was already working.
Hold rate above 45% on a 30 second DTC video is considered strong, and below 30% indicates the middle is losing people before the offer lands. TikTok benchmarks sit higher on both, since the platform's baseline scroll behaviour is different.
Delivery systems increasingly read the creative itself to decide who sees it. When the ad content is an input the optimiser uses rather than just something served to a chosen audience, creative metrics become leading indicators: hook rate this week predicts cost per acquisition next week, because the system is deciding distribution partly on engagement signals the creative produces.
Practically, that means a falling hook rate is an early warning rather than a retrospective observation. It moves first.
Use rates rather than counts, and normalise anything time-based by runtime.
The most common analytical error in creative testing is calling a winner too early.
After every test, write one sentence: what did this teach us that applies to the next ad?
If the answer is "variant B won", the test taught you nothing transferable. If it is "problem-first openings beat curiosity openings for this audience, twice now", you have learned something that makes every future ad better.
Most creative programmes never accumulate. They run tests indefinitely and start each one from scratch, which is a much more expensive way to be at the same place a year later.
Five, read in order: hook rate, hold rate, outbound click rate, frequency and cost per acquisition. Each maps to a different part of the ad, which is what makes them diagnostic. Cost per acquisition alone tells you an ad is bad without telling you which part.
30% or above on Meta, measured as 3-second video plays divided by impressions. TikTok benchmarks run higher, with strong performance in the 35 to 45% range, because baseline scroll behaviour on the platform differs.
The share of people who stayed after the first three seconds. Above 45% on a 30 second DTC video is strong, and below 30% means the middle of the video is losing people before the offer lands. It is the metric that tells you whether the hook wrote a cheque the body could not cash.
The opening works and the ad does not deliver on it. Usually the hook promised something the body never paid off, or the demonstration arrives too late. Keep the opening and rebuild the middle, rather than starting the ad over.
Hook rate is readable at a few thousand impressions, hold rate at a few hundred 3-second plays, outbound clicks at thousands of impressions, and conversion comparisons need roughly 30 to 50 events per variant. Judge creative on hook and hold; do not use conversion data to choose between two hooks.
