The short answer
Four creative metrics cover most decisions: hook rate (did they stop), hold rate (did they stay), CTR (did they act) and CPA (did it pay). Read them as a chain, not one at a time. A 2026 benchmark of 88,329 Meta video ads put the average hook rate at 25.4%, and on our board, videos between 31 and 120 seconds reach the top tier most often (about 10%).
The four metrics and how to calculate them
Most accounts drown in columns. For creative decisions you need four, plus two supporting ones.
| Metric | Formula (common version) | What it tells you | Belongs to |
|---|---|---|---|
| Hook rate | 3-second video plays / impressions | Did the first seconds stop the scroll? | Opener, first frame |
| Hold rate | ThruPlays / 3-second plays | Did people who stopped keep watching? | Body, pacing |
| CTR (link) | Link clicks / impressions | Did the ad make people act? | Offer, CTA, overall pull |
| CPA | Spend / purchases (or leads) | Did it pay? | The whole chain, including the landing page |
| Frequency | Impressions / reach | How often the same person sees it | Fatigue watch |
| CPM | Spend / 1,000 impressions | What the auction charges for this ad | Context only |
Hook rate and hold rate are practitioner metrics. Meta reports the inputs (3-second plays, ThruPlays, impressions) and you build the ratio as a custom column. ThruPlay is Meta's count of video plays to completion or to at least 15 seconds. Some teams calculate hold rate against impressions instead of 3-second plays. Either works if you stay consistent.
For images, hook and hold rate don't apply. Use CTR and CPA, plus outbound CTR if you have it.
What "good" looks like (benchmarks are directional)
A 2026 report from Billo, presented by Sovran, covers 88,329 Meta video ads with a sales goal and at least 1,000 impressions each, from January to June 2026. The average hook rate was 25.44%. The report itself frames these as reference points, not targets, and doesn't publish margins of error.
Use benchmarks to catch outliers, not to grade ads. A pet supplement ad and a toy ad don't compete for the same attention. Your own account median, over the last 90 days, is the better baseline.
Read them as a chain
Each metric is a gate. A viewer passes the hook, then the hold, then the click, then the purchase. Where the drop happens tells you what to fix.
| Hook rate | Hold rate | CTR | CPA | Diagnosis | What to change |
|---|---|---|---|---|---|
| Low | High | Fine | Good | Winner held back by its opener | New openers on the same body |
| High | Low | Low | Bad | Opener promises something the body doesn't deliver | Rewrite seconds 3 to 15, keep the opener |
| High | High | Low | Bad | People watch but aren't moved to act | Stronger offer, clearer CTA, proof earlier |
| High | High | High | Bad | Problem is after the click | Landing page, price, message match |
| Low | Low | Low | Bad | Concept doesn't land | Kill and move on |
| Fine | Fine | Fine | Good, then rising | Fatigue | Check frequency trend; add new hooks |
The first row is the most valuable one and the most often missed. A low-hook ad that still converts is a winner that only reaches a small slice of the audience. A better opener can widen it. For how to iterate openers, see iterations vs new concepts.
The fourth row catches a lot of teams too. High CTR with bad CPA usually means the ad and the page tell different stories. Our message match guide covers how to line them up.
Length changes how you read hold rate
A 15-second ad and a 5-minute ad can't be judged on the same hold rate. On the Ad Radar board (October 9, 2026: 2,443 live Meta ads above $50k in estimated spend), 1,897 videos have a known length. Here is how they spread:
| Video length | Ads | Share reaching top tier | Median est. spend | Median days live |
|---|---|---|---|---|
| Up to 15s | 109 | 4.6% | $236K | 98 |
| 16 to 30s | 184 | 8.7% | $192K | 110 |
| 31 to 60s | 613 | 10.1% | $263K | 123 |
| 61 to 120s | 467 | 10.3% | $247K | 121 |
| 121 to 300s | 420 | 8.3% | $224K | 122 |
| Over 300s | 104 | 5.8% | $193K | 100 |
"Top tier" is the top 10% of the board by Ad Radar's winner score. The middle bands (31 seconds to 2 minutes) reach it most often, and 524 winning videos run longer than two minutes. Long ads are common among winners. They just need a different yardstick.
64s
median length of a winning video
524
winning videos longer than 2 minutes
10.3%
top-tier rate for 61 to 120s videos
4.6%
top-tier rate for videos up to 15s
For a long ad, the ThruPlay mark (15 seconds) is early. Add a custom milestone, such as video plays at 50% or 75%, and judge the ad on how many people reach the offer. For a 10-second ad, hook rate and CTR carry almost all the signal. More in best video ad length.
These are not the project.
- Est. spend
- $1.4M
- Days live
- 56
- Format
- Video, 796s
A 13-minute video in the top tier after 56 days. Hold rate to 15 seconds says almost nothing here. Retention at the point where the offer appears is the metric that matters.
The first real innovation in toothpaste since 1914.
- Est. spend
- $668K
- Days live
- 98
- Format
- Video, 8s
- Variants
- 3
Eight seconds long. For an ad this short, the opener is nearly the whole ad, so hook rate and CTR carry the signal and hold rate barely matters.
Metrics you can't see on competitor ads (and what to use instead)
When you study competitors, none of these four metrics is public. The Meta Ad Library gives you three signals that stand in for them:
- Days live. A rough proxy for CPA. Brands rarely keep paying for an ad that loses money for months. The median ad on our board has been live 120 days.
- Variants. A proxy for confidence. Brands add versions to ads that work. Ads with five or more variants reach our top tier 57% of the time, against 4.5% for single-version ads (variants also feed the winner score, so the link is partly built in).
- Engagement over time. The base for spend estimates. Ad Radar models estimated spend from public engagement tracked over time, which is useful for ranking ads against each other, not for exact budgets.
These won't tell you an ad's hook rate, but they tell you whether it survived. That's the question that matters when you pick references for a brief.
This is how your foot was born.
- Est. spend
- $1.5M
- Days live
- 118
- Format
- Video, 293s
- Variants
- 2
Nearly five minutes, an unaware opener and a listicle landing page. Nobody judging this ad on 3-second plays alone would know why it reached the top tier. The public signals (118 days, two variants, seven-figure estimated spend) say it holds.
A creative metrics dashboard you can build in Ads Manager
CUSTOM COLUMNS (Ads Manager > Columns > Customize > Create custom metric)
Hook rate = 3-second video plays / Impressions
Hold rate = ThruPlays / 3-second video plays
Offer reach = Video plays at 75% / 3-second video plays (long videos)
COLUMN SET "Creative review" (ad level, last 7 days, compare to last 90 days)
Ad name | Spend | Impressions | Hook rate | Hold rate | CTR (link) |
CPA | Frequency | Days since launch
RULES OF THUMB (adjust to your account baseline)
- Judge hook rate after ~1,000+ impressions; CPA after 2-3x target CPA spend
- Flag hook rate < 0.8x account median -> opener test
- Flag CTR high + CPA > 1.3x target -> landing page / offer check
- Flag frequency up + CTR down 2 weeks -> fatigue, add new hooksBring this column set to the weekly creative review and read each row left to right, using the diagnosis grid above. Most weeks, two or three rows explain where the next round of briefs should go. For the testing rules these metrics plug into, see the creative testing framework and our hook rate benchmarks.
Figures marked as estimated spend come from Ad Radar's model of engagement on public Meta Ad Library ads. They are estimates, labeled as such, and are best used to rank ads against each other.