The first three seconds: a working theory of ad hooks
Most video ads win or lose before the message even starts. This is how we measure the opening, the eight hook patterns that keep working, and how we test them without guessing.
People don't really watch your ad. They watch the first moment of it and decide whether the rest is worth their time. That decision is fast and mostly unconscious: an attention study of video ads found that while the average ad stays on screen for 12.3 seconds, viewers actually pay attention for just 3.7 of them.¹ TikTok's own research is even more blunt: 50% of an ad's total impact lands in the first 2 seconds, and the first 6 seconds capture roughly 90% of its cumulative effect on ad recall.²
So for most of the people you pay to reach, the hook is the only part of the creative they will ever see. This article is the working theory we use at Adside: how to measure hooks, why most of them fail, which patterns keep winning, and how to test them so the answer comes from data instead of taste.
The short version
- Measure the opening with hook rate (3-second views ÷ impressions): 20–25% is solid on Meta, 30%+ is top tier. Pair it with hold rate so you know the video keeps the attention the hook earned.
- Hooks usually fail the same three ways: opening on a logo, explaining context before there's any tension, and pacing the video like the viewer already agreed to watch.
- Winning hooks skew negative and declarative. In one analysis of 550+ proven winners, 91% opened with something negative and 82% opened with a statement rather than a question.
- A hook that overpromises will hurt you. It juices hook rate and quietly kills conversion downstream, so don't ship an opening the body can't back up.
- Test 3–5 hooks per concept on an identical body. The hook swap is the cheapest creative change you can make, and usually the one with the biggest payoff.
What a good hook rate looks like
Hook rate is 3-second video views divided by impressions. It answers one question: of everyone your ad was served to, what share stopped scrolling long enough to register it? Some teams call the same calculation thumb-stop ratio. Either way, it's a custom metric you build from two native fields, since the platforms don't report it directly.
The working benchmarks on Meta: a hook rate of 20–25% is a solid target range, and top performers exceed 30%.³ Placement matters more than most teams account for. Reels and Stories typically support higher rates than Facebook Feed for equivalent creative, with roughly 30–35% strong on Reels versus 20–25% solid in Feed, where clutter and muted autoplay drag everything down.⁴ So compare hooks within a placement, never across it.
Hook rate alone, though, only tells you people paused. The companion metric is hold rate (ThruPlays divided by 3-second views), which measures whether the video kept the attention the hook earned. A 40–50% hold rate is the usual target, with strong creative exceeding 50%.³ A great hook on a weak body shows up exactly here: high hook rate, collapsing hold rate. Read them together, otherwise you end up optimizing for a pause instead of an actual viewer.
Why hooks fail
The feed doesn't owe you any attention. Every swipe is a fresh auction for it, and unless your first frame gives someone a reason to stop, the default outcome is the next swipe. Most weak hooks fail in one of three ways:
Opening with the logo. This is a habit imported from TV, where the viewer couldn't leave. Early branding does help recall, but it should ride along inside the hook (product in hand, a watermark, a spoken mention) rather than be the hook. A title card spends three seconds of your most expensive real estate announcing that an ad is starting.
Slow builds. Establishing shots, ambient intros, a creator saying "hey guys, so today I want to talk about…". All of it assumes the viewer has already agreed to watch, and they haven't. That agreement is exactly what the hook is negotiating. If your story needs setup, you've sequenced it wrong: start at the moment of highest tension and backfill later.
Context before conflict. "We're a family-owned skincare company founded in—" is context. "Your moisturizer is why your skin is dry" is conflict. People stop for tension, surprise, or self-recognition; they do not stop for background information. Give them the conflict first and they'll sit through the context afterwards.
Before you brief any video, write the first three seconds as their own deliverable. If the opening can't survive on its own in a feed, the rest of the script never gets seen anyway.
Eight hook patterns that keep working
These are patterns rather than scripts. Each one works because of a mechanism, and the mechanism is the part you should copy. Before you pick one, know the skew: when one team analyzed 550+ proven winning ads, 91% had something negative in the hook (a warning, a mistake, a problem) and 82% opened with a statement rather than a question; 86% relied on a visual hook rather than a talking head alone.⁵
| Pattern | Example opening | Why it works |
|---|---|---|
| Negative hook | "Stop using retinol like this, it's aging you faster." | Brains prioritize threats over promises, which is why 91% of winners skew negative. |
| Unexpected visual | The product frozen inside a block of ice, hammer raised. | The feed trains people to predict what's coming; a frame that breaks the pattern forces a second look. |
| Direct callout | "If you manage ad accounts for more than three clients…" | Self-selection: the viewer hears their own description and grants relevance instantly. |
| Mid-action start | Open on the demo already halfway done, no setup. | Starts at peak motion and skips the part people skip. Movement itself stops thumbs. |
| Claim + receipt | "We cut their CPA 38%, and this is the actual dashboard." | Pairing the claim with on-screen proof disarms the skepticism reflex before it fires. |
| Question implying a gap | "Why do your best ads die on day six?" | Opens an information gap the viewer wants closed. Use it sparingly, since statements win more often. |
| Before/after flash | One second of the end result, then "here's how." | Sells the destination first, so the rest of the video becomes the route there. |
| "Watch me do X" | "Watch me build a landing page in 90 seconds." | Process is inherently watchable, and the time-box promises the payoff is near. |
Two platform notes before you go build. On TikTok, put a human in the frame immediately: TikTok's creative research found that showing a person in the first two seconds increases hooking power by about 50% and lifts ad recognition by 32%.⁶ And on Meta Feed, assume silence. The hook has to land with text overlay and image alone, because much of Feed plays muted.
The hook has to keep its promise
Teams that get good at hooks run into a specific failure mode: the hooks get better than the product story can support. A shock opening or an inflated claim will absolutely raise hook rate, and then every metric downstream of it decays. Viewers who were stopped by a promise the body doesn't keep don't click; the ones who click anyway don't convert; and the ones who convert on a misrepresented promise refund. You've paid to disappoint people at scale.
What keeps this in check is coherence. The hook makes a promise, the body proves it, and the landing page collects on it. Before shipping a hook, ask one question: if someone watched only these three seconds and then converted, would they get what they expected? If not, the hook is borrowing performance from the rest of the funnel.
If the hook promises something the body and the landing page don't deliver, hook rate stops meaning anything. Check the whole chain before you celebrate the number.
This is also why the negative-hook statistic shouldn't be read as "be alarmist." The winners open on a real problem the product actually solves. Negativity that resolves into relief converts. Negativity that resolves into nothing is just clickbait with better production values.
Testing hooks systematically
Hooks are the best place to spend testing budget in video creative because they're the cheapest thing to change. You don't reshoot anything: the body, offer and CTA stay identical and only the first three seconds vary. That makes the hook swap the first rung of any refresh ladder and the fastest way to turn one good concept into a portfolio.
The protocol that keeps results readable:
- 3–5 hooks per concept, identical bodies. Fewer than three and you're measuring luck; more than five splits delivery too thin to learn anything at normal budgets.
- Make the hooks mechanically different rather than cosmetically different. A negative hook vs. a direct callout vs. a before/after flash teaches you something; three line readings of the same sentence don't.
- Judge per placement, against your own account baseline, after each variant has a few thousand impressions.
- Keep the losers' learnings. A hook that failed for this concept is still data about your audience, so write down why you think it lost.
Producing those variants used to be the bottleneck, and it's the step we've automated at Adside: the Video Ad Generator renders multiple hook variants from one script or URL, and the Creative Performance Predictor scores them before any spend, so you know what order to test them in.
Then read the metrics as a system, because each combination points to a different fix:
| Hook rate | CTR | CVR | Diagnosis |
|---|---|---|---|
| High | High | High | A winner. Scale it and queue its successor now, because this is the ad fatigue will hit first. |
| High | Low | — | The hook stops people; the body or CTA doesn't move them. Fix the middle, keep the opening. |
| High | High | Low | Coherence break: the promise doesn't match the landing page. That's a message problem rather than a creative one. |
| Low | High | High | The few who stop do convert, so the hook is the bottleneck. Swap hooks before you kill the concept. |
| Low | Low | — | The concept itself isn't landing. No hook will save it; you need a new angle. |
The fourth row is the one teams miss most. A low hook rate with strong downstream metrics is a good ad wearing a bad opening, and killing it wholesale throws away a proven body that three new hooks might have rescued.
Hooks beyond video
The three-second logic applies to anything that lives in a feed, not just video. Every format has a hook (the element that gets processed before the scroll decision) and the same patterns transfer:
- Primary text: the first line before the fold is your hook. "Stop paying for clicks that never convert" earns the "see more" tap; "At [Brand], we believe…" does not.
- Static ads: the headline-visual pair does the whole job at once. A static is basically a hook with nowhere to hide, which is why the pattern table above works as a headline-writing checklist.
- Carousels: card one is the hook and earns the swipe; the payoff lives on cards two through five. Spending your best material on card one's body copy is burying the lede.
- UGC: the first sentence out of the creator's mouth is the hook, and it's where most UGC briefs are vaguest. Scripting five first-lines per brief (or pressure-testing them with a UGC Simulator before production) costs minutes and decides most of the outcome.
Treat the hook as a unit of creative work in its own right, something you write and test separately from the asset it opens, and every format in the account gets cheaper to improve.
Frequently asked questions
What is a good hook rate?
Hook rate is 3-second video views divided by impressions. On Meta, 20–25% is a solid working range and top performers exceed 30%. Benchmarks shift by placement (Reels and Stories typically run higher than Facebook Feed for the same creative), so build a per-placement baseline from your own account rather than chasing a single global number.
How many hooks should I test per concept?
Three to five. Fewer than three and you are mostly measuring luck; more than five splits delivery so thin that nothing reaches significance on a normal budget. Keep the body and offer identical, vary only the first three seconds, and kill the losers once each variant has a few thousand impressions.
Do hooks differ on TikTok vs Meta?
The principles transfer, but the pacing doesn't. TikTok's own research shows 50% of an ad's total impact lands in the first 2 seconds, so TikTok hooks need to be faster, native-looking and usually creator-led; showing a person in the first two seconds measurably increases hooking power. Meta tolerates slightly slower openings, especially in Feed, and text overlays carry more weight because much of Feed plays muted.
Is hook rate the same as thumb-stop ratio?
Effectively yes, both are usually computed as 3-second video views divided by impressions. Neither is a native platform metric, so check how your team or tool defines it before comparing numbers. The companion metric is hold rate (ThruPlays divided by 3-second views), which tells you whether the video keeps the attention the hook earned.
Should the brand appear in the first three seconds?
For recall, early branding helps; attention research consistently favors branding early rather than saving the logo for the end. But a logo is not a hook. Lead with the tension or the visual that stops the scroll, and let the brand ride along as a watermark, a product-in-hand, or a spoken mention rather than a title card.
Sources
- Attention vs. on-screen time for video ads (3.7s of 12.3s) — Marketing Week, Most B2B video ads fail to gain adequate attention or drive recall
- TikTok Marketing Science on impact in the first 2 and 6 seconds — Social Media Today, TikTok Shares New Notes on Maximizing Ad Effectiveness
- Hook rate and hold rate formulas and benchmarks — Vaizle, Hook Rate and Hold Rate: Facebook Ads Formulas and Benchmarks
- Thumb-stop ratio benchmarks by placement — Thumb Stop Ratio: Benchmarks, Formula & Fixes
- Analysis of 550+ proven winning Meta video ads — The Performers, The Formula for Proven Winning Video Ads
- Showing a person in the first 2 seconds (+50% hooking power, +32% ad recognition) — TikTok × CreatorIQ, Data-Backed Keys to Success for Advertisers