How to Read a Hook Rate Report and Act on It

Hook rate is 3-second video views divided by impressions. Pull it at ad level, in the same attribution window you use for the rest of the test. Do not mix thruPlay into that fraction — thruPlay is a longer-view metric and will make every open look weaker than it is. Campaign totals hide a dead file behind a strong one. The formula is simple; the error is almost always the column.

On Meta, use 3-second video plays over impressions. Exclude a placement that distorts the mix if you have to — Audience Network is the usual culprit — and do not average prospecting with retargeting in one row. ThruPlay is closer to hold than to hook: it counts 15-second views, or completion if the video is shorter. Using thruPlay over impressions as a 'hook rate' will push you to rewrite the middle of the video when the actual problem is the opening, or the reverse. Keep two ratios on the sheet: hook as 3-second views over impressions, hold as 15-second views over 3-second views.

Wait for a directional read: 48–72 hours or about 1,000 impressions per variant, whichever comes first. Earlier than that, delivery is still bouncing between people and placements, and a 12% versus 40% 'winner' is often who got the first Reels burst. You are ranking ads in an auction, not publishing a paper. About 1,000 impressions is a diagnostic threshold, not a p-value. Anyone promising a scientifically proven hook from one afternoon of Reels is selling theatre, and a calculator on 11 clicks is worse than a simple rule.

Retargeting hook rate looks better because the viewer already knows the brand, the product, or the face. Recognition is a thumbstop. A 45% hook on a 1% conversion audience does not mean the creative would do that on cold traffic, and using it as your benchmark will make every prospecting ad look like a failure. Judge new creative on the cold number.

Split reporting every time: cold versus view, engagement, or purchaser retargeting. Use the retargeting lift as a consistency check — if the same avatar keeps winning on warm traffic, that is a casting signal you can reuse, not proof that the hook is globally strong. Warm numbers are allowed to be higher. They are not allowed to set the floor for a prospecting test. If 80% of the spend in the row is retargeting, the 36% hook rate is a warmth number wearing a campaign badge, and every new file will look like a failure next to it.

Reels and Feed are not the same unit either. Reels is full-screen, sound-optional, swipe-vertical, and closer to TikTok than to a static Feed placement. A creative can hook at 34% in Reels and look ordinary in Feed because the unit itself supplies motion and face size. Always break hook rate out by placement before you declare a winner. Build 9:16 first. If you need Feed, export a 1:1 or 4:5 cut rather than letterboxing, and judge that cut on Feed stats, not on the Reels halo that was carrying the campaign number.

Never read hook rate alone. Target 30%+ on cold traffic; below about 20% the opening is weak. Hold should sit in a healthy 12–20% band. Outbound CTR should clear 1%+ on cold. The three numbers together tell you whether people looked, whether they stayed, and whether they wanted the click — which is the only version of 'the hook worked' that is safe to scale.

A 35% hook with a 6% hold means people looked and left. A 22% hook with a 16% hold and a 1%+ CTR can still be a quieter winner on cold traffic. A 40% hook with hold well under 12% is a false thumbstop: clickbait faces, fake arguments, and shocking first frames manufacture the pause and destroy CPA. Autoplay can also credit a 3-second view when the viewer never really looked, which is why hook can look healthy while comments, outbound CTR and hold look dead. If hook is 32% and hold is 5%, believe the hold. If hook is 32%, hold is 15%, and outbound CTR is 1.2% on cold, believe the hook.

Do not optimise the campaign for hook rate. Optimise for the business event — purchase, lead, add to cart — and use hook as a diagnostic, not a delivery goal. Meta will happily find people who watch three seconds if you ask it to; those people are not automatically buyers. Hook rate tells you whether the creative has earned the right to be judged on CPA. It does not replace CPA. Vanity optimisation is how accounts fill with entertaining ads that do not pay the bills, and it is the most common misuse of a healthy hook-rate column.

Weak hook with decent hold and CTR on the people who stay is a hook problem. Strong hook with weak hold is a mismatch between the opening promise and the next ten seconds. Strong hook and hold with weak CTR is a creative-to-offer or creative-to-page problem. Strong click and terrible CPA is usually offer, landing page, or tracking. Fix in that order: hook, body, click, offer.

Casting shows up as a one-face pattern. If three hooks all die on avatar A and live on avatar B, the face, age band or setting is wrong for that audience, not the copy. That is why the starting batch is six creatives — 3 hooks × 2 avatars on one body script. Two avatars is enough to separate casting from copy without slicing budget into noise. Do not average them. Kill the weak face, keep the winning face, and write more hooks against it. Preview taste is a terrible casting tool; the cold hook rate is the tool.

Frequency can make a good hook look dead. On cold audiences, creative fatigue typically shows at frequency 2.5–3.5: hook rate and CTR slip together, CPA rises, and the same opening that used to stop the thumb becomes wallpaper. Confirm by checking whether new people still hook. The cheap refresh is a new first line on the same body and avatar. Rebuilding the whole ad because frequency cooked a winner is the most common waste in a creative team. If hook and CTR are stable and blended revenue fell across every channel, look at tracking or demand next — not the first frame.

A new first two seconds is the fix, not a new music bed, a new colour grade, or a longer intro 'for context'. Start talking or showing the problem in the first sentence, put a face and on-screen text on frame one, cut any logo sting, and write a line a real person would say out loud. If you cannot explain the video's job in one spoken clause before second two, the hook is not finished.

Start with three families, not fifteen: a problem call-out, a pattern interrupt, and a specific proof or demonstration. Those three cover most of the work a 9:16 open has to do. Question hooks and wait-don't-scroll lines are overused and often read as ads instantly. Native-sounding, specific, slightly unfinished lines beat slogans. Keep them under 15 words. On-screen text in the first two seconds usually helps because a large share of feed viewing starts muted — a short, high-contrast line that matches the spoken hook, inside the safe zone, under about seven words. Mismatched text burns trust in the same two seconds you are trying to win.

A more realistic avatar will not automatically raise hook rate. Realism helps only when the current face is the reason people bounce — obvious artefacts, wrong age band, or a setting that screams studio. A slightly more realistic avatar delivering a slogan will still lose to a merely decent avatar delivering a specific, human line. Casting and copy both sit in the first two seconds; copy is usually the cheaper test. Run the 3 × 2 batch before you commission a custom clone. If both stock faces hook under 20% on cold with good scripts, then you have a casting or offer problem, not a render-quality problem.

Lock the body script, the proof beats and the CTA. Only regenerate the opening scene. Put three new hooks against two avatars, launch on the conversion objective, and read hook rate again at 48–72 hours or about 1,000 impressions per variant. Mixing hook and avatar changes on a small budget is how you learn nothing. Once a family wins, iterate inside that family rather than throwing eight unrelated concepts into the same set.

Three hooks against two avatars on one body is the starting batch because it separates copy from casting without shredding budget. More than about six to eight variants in one ad set on a modest daily budget turns the test into noise: nothing reaches about 1,000 impressions in 48–72 hours. If you already have a winning face, 3–5 new hooks against that face is a valid second test. If you already have a winning hook, two new avatars is enough. Volume is useful; simultaneous volume is not the same as learning.

Klip Kanvas is built for that iteration loop — new hook, same body — so you can ship the six-file batch in one sitting instead of a new shoot. Name each file so the next hook-rate report still reads without a spreadsheet archaeology session. Then still watch the winner on mute, 9:16, at arm's length. Editor previews measure taste; hook rate measures a cold thumb on a phone. If the team favourite dies and the slightly messy one hooks at 30%+, believe the feed, not the Slack thread that voted on previews.