Hook Rate: What It Measures and How to Improve It
Hook rate is industry shorthand for how often a video ad holds a viewer past its first seconds. How to calculate it from documented metrics, read it and test better hooks.
02 / In this article
Hook rate is industry shorthand for the share of impressions in which a video ad was watched past its first few seconds. It is not a metric you will find in Meta, TikTok or Google Ads under that name, so you calculate it from documented metrics, and it only means something when you compare it between variants of the same ad. This guide shows how to build it, what it can and cannot tell you, how it differs from hold rate, and how to write and test better hooks.
Key Takeaways
- Hook rate is a ratio you build yourself. The usual recipe divides an early-play metric (Meta's 3-second video plays, TikTok's 2-second video views) by impressions. No platform publishes it under that name.
- It measures stopping power, not persuasion. A high hook rate says people stayed for the opening. It says nothing about whether they understood the offer or bought.
- Placement, autoplay and format change the number. Each platform counts a "play" differently, so compare only like with like: same platform, same placement, same period.
- Hook rate and hold rate answer different questions. Hook rate is about the first seconds. Hold rate is about whether people stay after the opening.
- Test hooks as variants of one ad. Change only the opening, keep the rest identical, and read the result as a relative difference, never as a pass or fail against a universal number.
What is hook rate?
Hook rate is the proportion of ad impressions that turn into an early video play, used as a signal of whether the opening of a video ad stops the scroll. A "hook" is the first moment of an ad, the part that has to earn the next second of attention before anything else can happen.
The term comes from practitioners. Meta's Business Help Center, TikTok's Ads Manager glossary and the Google Ads Help Center each document their own video metrics, and none of them lists a metric called "hook rate" (as of October 2026). That matters: two people who say "our hook rate is high" may have divided different numbers, and any figure quoted as "good" is someone's habit, not a platform standard. This article gives no thresholds for that reason. If you want the wider picture of how this fits into planning and testing, start with our guides to creative testing and creative analysis.
How do you calculate hook rate from documented metrics?
You calculate hook rate by dividing the platform's early-play count by the number of impressions, using the platform's own metric names.
The formula in plain words:
Hook rate = early video plays ÷ impressions
What counts as an "early video play" depends on the platform. Here is what each one documents.
Meta (Facebook and Instagram). The metric is called 3-second video plays. Meta defines it as the number of times your video played for at least three seconds, or for nearly its total length if it is shorter than three seconds. For short videos, the help article says the threshold is 97% of total length. Plays are counted separately for each impression, and time spent replaying the video is excluded. Meta also states that it calculates this metric with a "unique seconds" method: it sums the distinct seconds of a video that played, so replayed segments do not add up. So for Meta, hook rate is 3-second video plays divided by impressions.
TikTok. The Ads Manager glossary lists 2-second video views and 6-second video views. A 2-second view is counted when the video starts playing for at least two seconds in an impression session, with replays excluded. A 6-second view works the same way, with two additions: a video shorter than six seconds counts if it plays in full, and an engagement in the first six seconds also counts. For an early-attention ratio, 2-second video views divided by impressions is the closest match to Meta's metric. TikTok's glossary also says that if a viewer swipes away and then back, that is counted as two impressions, which affects the denominator.
YouTube and Google Ads. The help article "About YouTube ads and view metrics" does not document a metric for the first few seconds. It documents TrueView views, which for a skippable in-stream ad are counted when a viewer watches 30 seconds, or to the end if shorter, or clicks an element of the ad. For YouTube Shorts ads the threshold is 10 seconds, or a click on the call to action, "Like", "Comments" or "Share". Its "Video played to" metrics (25%, 50%, 75%, 100%) use viewers who started the player as the base, not all impressions. So on Google's side there is no like-for-like hook rate: you can build a view-based ratio, but it measures something later and different. Label it accordingly.
Write the formula next to the number in every report. "Hook rate (Meta: 3-second video plays ÷ impressions)" can be checked. "Hook rate: high" cannot.
What does hook rate actually tell you?
Hook rate tells you how often the opening of a video earned a few more seconds of attention, compared with other openings shown in similar conditions. It is a measure of stopping power.
That makes it useful for one decision: which opening to keep. It is not a measure of creative quality as a whole. An ad can stop thumbs and then lose them, or stop fewer people and convert the ones who stay better.
Think of hook rate as a diagnostic, not a goal. The business result sits further down the chain: clicks, conversions and revenue, which you measure with your analytics and tracking setup. Optimising only the top ratio tends to produce clickbait openings, so read it alongside the later metrics.
What is the difference between hook rate and hold rate?
Hook rate covers the first seconds of a video, while hold rate covers how well the video keeps the people who stayed past the opening. Both are practitioner terms, and hold rate has no single definition.
A common version of hold rate divides a longer-view metric by the early-play metric. On Meta that could be ThruPlays divided by 3-second video plays. Another version divides ThruPlays by impressions. Both exist in the wild, so pick one, write it down, and never mix them in a single comparison.
Meta documents ThruPlay as the number of times a video was played to completion or for at least 15 seconds. For videos shorter than the set duration, it counts after at least 97% of the video was played, and replays within a single impression are excluded. If you want a watch-time view, Meta also documents video average play time: the time spent playing the video, including replays for a single impression, divided by the total number of video plays. Because it includes replays, Meta notes the reported average could be longer than the video itself.
Meta also reports milestone metrics such as video plays at 25%, 50%, 75%, 95% and 100%, which count people who reached those points "whether played continuously or skipped to those points". That differs from continuous watch time, so do not use them as a stand-in for hold rate without saying so.
| Question | Metric family | Where the ratio sits |
|---|---|---|
| Did the opening stop the scroll? | Hook rate | Early plays ÷ impressions |
| Did people stay after the opening? | Hold rate | Longer plays ÷ early plays (or impressions) |
| Did the middle and end earn attention? | Milestones, average play time | Reached points, time per play |
| Did the ad lead to action? | Clicks, conversions | Outcomes per impression or per play |
The table in the figure reflects only what the help pages above state, as of October 2026. Where a cell says "not stated", we found no statement in the pages we read.
Need more openings to test?
Send us one ad and the offer behind it. We reply with how we would produce a set of hook variants for it.
See AI video adsWhy can the same hook score differently in different places?
The same video can produce a different hook rate depending on where and how it plays, because each platform decides when a video starts and what counts as a play.
Meta documents a clear example. For a video ad to trigger autoplay or click-to-play, a minimum percentage of its pixels must be in view, and the percentage depends on the placement. As of October 2026, the Meta help page lists 50% for Facebook Feed and Facebook Reels, 25% for Instagram Feed and Instagram Explore, and 100% for Facebook Stories and Instagram Stories. So a hook that starts when 25% of the ad is visible in one placement begins later in a different one. If you then pool placements, the ratio you see is a blend of different viewing situations.
Other differences work the same way. Google documents that an in-feed impression is counted when the viewer sees the thumbnail, an in-stream impression when the video starts, and a Shorts impression when playback starts in the Shorts feed. TikTok counts a swipe away and back as two impressions, while Google says the same move on a Shorts ad is not an additional impression or view. And viewers can watch with sound off, which no platform reports, so a hook that depends on a spoken line may behave differently from one that works in silence.
The practical rule is simple. Compare hook rates only within the same platform, the same placement set, the same objective and the same period. When in doubt, break the result out by placement before drawing a conclusion.
How do you diagnose a weak hook?
You diagnose a weak hook by comparing the early-play ratio against the ads around it, then checking the metrics on either side to find where the drop happens.
Start with the three ratios side by side: hook rate, hold rate and a click or conversion measure. Each pattern points to a different place to look.
Read the four corners like this:
- Low hook rate, healthy hold rate. The people who stay like the video, but too few start. The opening is the problem. Test new first seconds on the same body.
- High hook rate, low hold rate. The opening stops people and the next beat loses them. Check whether the hook over-promises or the move from hook to message is slow.
- Both low. The concept itself may be off for this audience, or the video is mismatched to the placement. Changing the opening alone is unlikely to fix it.
- Both healthy, weak clicks or conversions. Attention is not the constraint. Look at the offer, the call to action, the landing page and the tracking before touching the hook again.
Before you act on any of this, check that both variants ran in the same placements and period, that their lengths are comparable, and that the counts are large enough for the gap to mean something.
The "enough data" question has a statistical answer. The NIST/SEMATECH e-Handbook of Statistical Methods describes how to test whether two processes have the same proportion, using a z statistic for large samples and an exact test for small ones. A hook rate is a proportion, so that logic applies directly. You do not need to run the maths by hand, but you should know that a small gap on small counts is exactly the case where the test says "cannot tell".
How do you write a better hook?
You write a better hook by deciding what the viewer should understand or feel in the first seconds, and then making that visible and audible in the opening frame instead of building up to it.
The first seconds of a video ad do three jobs: they signal that the ad is relevant to this person, they create a reason to keep watching, and they avoid anything that looks like a pause. The opening has to work with the sound off.
Here are opening types that practitioners commonly use, framed as options to test, not as rules:
- Result first. Open on the outcome the product or service leads to, then show how. Works when the outcome is visual and specific.
- Problem first. Open on the irritation the viewer already knows, in their words. Works when the problem is recognised quickly.
- Question. Ask something the target viewer would answer in their head. Works when the question is narrow, not generic.
- Pattern interrupt. Start on something unexpected in the frame or the action. It has to connect to the message, or it becomes a trick the rest of the video cannot support.
- Direct address to a person. Someone speaks to the camera about the situation. This is the opening style most associated with creator-style ads, and our UGC ads service is built around that format.
- Demonstration. Show the product doing the job in the first frame.
Whichever you choose, treat the first three seconds as three beats: the subject in frame one (close and clear, no logo sequence), the tension in seconds one to two (the problem, result or question), and a bridge at second three that gives the viewer a reason to stay past the early-play mark.
Add on-screen text that carries the idea without audio, and check the opening on a phone in a muted setting before you approve it. Keep the claim honest: an overstated hook creates a mismatch with the landing page.
Turn one concept into several hooks
We direct AI-assisted production of hook variants from your brief and images, so your ad accounts have something to test.
Explore AI ad creativesHow do you test hooks properly?
You test hooks by producing several versions of the same ad that differ only in the opening, running them under the same conditions, and comparing the early-play ratio and the metrics after it.
Meta's help article on A/B testing is a good model for any platform. It says a test compares two versions by changing a variable such as an ad image, ad text, audience or placement, and that nobody sees both versions. It recommends choosing a hypothesis first, using the same budget for both versions, and avoiding informal tests such as switching ad sets on and off by hand, which can lead to unreliable results. Our A/B testing service applies the same discipline.
A hook test, step by step:
- Write the hypothesis. "An opening on the result will pass the first seconds more often than an opening on the product, for this audience."
- Hold everything else constant. Same body, same call to action, same audience, same placement set, same budget.
- Change only the first seconds. If you change the voice-over, the music, the cut and the text at once, you will not know which one mattered.
- Define the read-out in advance. Name the early-play ratio, the hold ratio and the outcome metric, with their formulas, before launch.
- Decide the stopping rule in advance. Pick the run length and the minimum counts you need, so you do not stop when a chart looks good.
- Read relative differences. The question is "which variant is ahead of the others, and by enough to trust", not "is this number good".
- Check the later metrics. A winner on the early ratio that loses on clicks or conversions is a lead worth following, not a result to scale. Record what the winning opening did.
Keep a naming scheme so every result traces back to the opening type, such as "result-first", "question" or "demo". That turns a pile of tests into a library, the basis of a longer creative strategy described in the creative strategy framework.
One more caution on frequency of change. Hooks wear out like any other creative element, because the same people see the same opening. When your ratio declines over time for a running ad, read it together with the signals described in our guide to creative fatigue before deciding whether the opening is weak or simply seen too often.
What do you do when a hook wins?
When a hook wins, keep the idea, not the file: reuse the opening logic across new versions, new formats and new audiences, and keep testing it against fresh challengers.
A winning opening usually carries a repeatable idea, such as "start on the result". Apply it to the next product, audience segment and format, then test again. Treat any win as provisional: it holds for the audience, placement and period in which you measured it.
Do not stretch a hook past what the product delivers. If the winning opening makes a claim, check that the claim is accurate and that the landing page continues it. A mismatch will show up in later metrics.
This is where production speed matters. Testing hooks properly means having several versions to run, and most teams are short of variants, not ideas. Versai's AI ad creatives work is built around AI-assisted production with human direction, so a single concept can be turned into a set of variants for your accounts. We direct the hooks, the framing and the cuts. We make no promise about how the numbers move: that depends on your offer, your audience and your accounts. For the video side specifically, see video ads built as sets of variations, and for planning the paid side around them, our paid media service.
The people who connect these tests to business decisions fill the role of a creative strategist.
Frequently Asked Questions
What is hook rate in video ads?
Hook rate is industry shorthand for the share of impressions in which a video ad was watched past its first seconds. It is built from documented metrics, such as 3-second video plays divided by impressions on Meta, and it is not an official platform metric.
What is a good hook rate?
There is no verified universal number. Platforms do not publish a target, and the figure changes with placement, audience, format and how you calculate it. Use relative comparison instead: which of your own variants is ahead under the same conditions.
Is hook rate the same as hold rate?
No. Hook rate is about the opening seconds. Hold rate is about whether people stay after it, usually a longer-view metric such as ThruPlays divided by early plays or by impressions. State which formula you use.
How do you calculate hook rate on TikTok?
The closest match is 2-second video views divided by impressions. TikTok defines a 2-second view as a video starting to play for at least two seconds in an impression session, with replays excluded. Write the metric names beside the result.
Does Google Ads have a hook rate metric?
Not in the help material we read. Google documents TrueView views, view thresholds that depend on the ad format, and "Video played to" metrics based on viewers who started the player. Those measure later behaviour on a different base, so label any ratio you build accordingly.
Can a good hook fix a weak offer?
No. A hook can win the first seconds, but the offer, the call to action and the page decide what happens next. If hook rate is healthy and conversions are weak, look downstream.
How many hooks should I test at once?
Enough to compare clearly and few enough to read cleanly. A small set of openings on one identical body, under the same conditions, is easier to read than many mixed changes.
Next steps
Pick one running video ad, write down the formulas for hook rate and hold rate, and record a baseline per placement. Draft a hypothesis and produce three or four openings on the same body. Read the results against each other, not against a number from the internet. For help producing the variants or setting up the test, see our AI ad creatives service, the Meta Ads work it supports, or get in touch with the ad you would like to test.
Sources
- Meta Business Help Center, "3-second video plays": https://en-gb.facebook.com/business/help/743427195703387
- Meta Business Help Center, "ThruPlays": https://en-gb.facebook.com/business/help/471190536725647
- Meta Business Help Center, "Video average play time": https://en-gb.facebook.com/business/help/1794520550789165
- Meta Business Help Center, "About video metrics for ads": https://en-gb.facebook.com/business/help/2146677122218192
- Meta Business Help Center, "About video ad metrics calculation": https://en-gb.facebook.com/business/help/1868286323447328
- Meta Business Help Center, "Video ad playing specifications by placement": https://en-gb.facebook.com/business/help/2013114112289197
- Meta Business Help Center, "About A/B testing": https://en-gb.facebook.com/business/help/1738164643098669
- TikTok Ads Manager Help Center, "Video play": https://ads.tiktok.com/help/article/video-play?lang=en
- Google Ads Help, "About YouTube ads and view metrics": https://support.google.com/google-ads/answer/2375431
- NIST/SEMATECH e-Handbook of Statistical Methods, "How can we determine whether two processes produce the same proportion of defectives?": https://www.itl.nist.gov/div898/handbook/prc/section3/prc33.htm
Newsletter
One short email when we publish
Notes on AI, image, video and brand work, from the people who do it. No spam.