Meta hook rate benchmarks: formula, hold rate and what counts as good

The Meta metrics behind hook rate

Hook rate and hold rate are industry formulas built from Meta's standard video metrics. Triple Whale states it plainly: Ads Manager has no built-in hook rate column, so you create one under Columns, Customize columns, Create custom metric. These are the Meta Business Help Center definitions you build it from:

1. 01

   Impressions: the number of times your ads were on screen. A video does not need to start playing for an impression to count, which Meta notes differs from industry standards for video.

2. 02

   3-second video plays: the number of times your video played for at least 3 seconds, or for nearly its full length if shorter. It counts unique seconds watched, once per impression, and excludes replays.

3. 03

   ThruPlays: the number of times your video played to completion or for at least 15 seconds. Meta treats 97% of the length as complete. Replays within one impression are excluded.

4. 04

   2-second continuous video plays: plays of at least 2 continuous seconds, mostly with at least 50% of the video's pixels in view.

5. 05

   Video plays at 25%, 50%, 75%, 95% and 100%: milestone metrics that include plays that skipped to that point, so they are not directly comparable with the duration metrics.

Hook rate and hold rate formulas

Every source we checked (Motion, Triple Whale, Billo, Vaizle, AdManage, Jetfuel and Foreplay, which calls it thumb-stop rate) uses the same hook rate formula: hook rate = 3-second video plays ÷ impressions × 100. An ad with 10,000 impressions and 2,600 3-second plays has a hook rate of 26%.

Hold rate has no single formula, which is the main reason published hold-rate benchmarks disagree. Motion and Jetfuel divide 15-second video plays by 3-second plays. Billo and Vaizle divide ThruPlays by 3-second plays. Some teams divide ThruPlays by impressions instead, which gives a much smaller number. Triple Whale's advice is the useful rule: pick one, label it in every report, and never mix them.

Our hook rate calculator uses 3-second plays ÷ impressions for hook rate and ThruPlays ÷ 3-second plays for hold rate, and tests whether the difference between two ads is statistically significant.

Hook rates are not comparable across platforms. TikTok reports 2-second and 6-second views, not 3-second plays, so a TikTok hook rate is a different measurement. Compare TikTok ads with TikTok ads.

Published hook rate benchmarks, source by source

Here is what each named source says, with its basis. Only Billo's figure comes with a stated sample.

1. 01

   Billo, H1 2026 video ad benchmarks (September 25, 2026): 88,329 sales-objective Meta video ads, January–June 2026, 14 industries, about $122M in spend. Average hook rate 25.44%, up from 24.42% in the previous period. Highest categories: Toys & Games at 28.79% and Business & Industrial at 28.74%.

2. 02

   Motion, creative metrics guide (2026, no sample stated): target a 30–40% hook rate as a baseline. Thumb-stop tiers: under 25% needs work, 25–35% solid, over 35% strong.

3. 03

   Vaizle (January 19, 2026, no sample stated): aim for a 20–25% hook rate and a 40–50% hold rate; top performers exceed 30% and 50%.

4. 04

   AdManage (February 5, 2026, based on practitioner sources from 2024–2026): under 20% needs fixing, 20–25% workable, 25–35% solid, 45% or more exceptional.

5. 05

   Jetfuel (August 20, 2026), citing AdSights data: Meta feed cold-prospecting hook rate 18–28% with a median of about 23%, Reels 24–36% with a median of about 30%; hold rate 12–25% with a median of about 18%.

6. 06

   Triple Whale (October 12, 2026): there is no universal good hook rate. Compare each ad with your own account average.

Why the benchmarks disagree

Three things explain most of the spread. First, sample: the only large dataset (Billo, about 25%) sits below the 30–40% targets that most blogs repeat without a sample. A target is not an average. Second, placement: Jetfuel's cited figures put Reels hook rates several points above feed, so an account that spends mostly on Reels will look better on hook rate. Third, formula: the hold-rate figures range from a median of about 18% to 40–50% partly because the sources divide different numbers.

Some sources also contradict themselves on the same page, giving two different targets in two sections. Treat any single hook rate threshold as a rough guide, not a pass mark.

How to set your own hook rate benchmark

A benchmark from your own account controls for your audience, placements and offer, which no published average can.

1. 01

   Create the custom metric in Ads Manager: 3-second video plays divided by impressions. Create hold rate the same way and label which formula you used.

2. 02

   Export the last 30 to 90 days of video ads with at least a few thousand impressions each. Small ads swing wildly.

3. 03

   Take the median hook rate, not the mean, and split it by placement (feed vs Reels) if you buy both.

4. 04

   Judge new ads against that median. An ad well above it has a working opening; an ad below it needs a new first three seconds, not a new body.

5. 05

   When two openings look different, check the difference in the hook rate calculator before you call a winner.

How to raise hook rate

Hook rate measures only the first three seconds, so fix the first three seconds. Change one thing per variant: the first line, the first visual action, or the on-screen text. Our UGC hooks library lists openings by angle and industry, and a UGC script template keeps the body constant while you swap openings. Write the variants into a UGC brief so a creator or AI actor films all of them in one session.

This is where production speed matters. With FireplaceUGC you can render several openings on the same body with the same AI actor in Talking Video, so the test isolates the hook. Category-specific openings are on our supplements and mobile apps pages.

Watch hold rate alongside it. A shocking opening can raise hook rate and lower hold rate if the next seconds do not deliver what the hook promised. The opening that wins is the one that keeps both up.