AI ad breakdowns
Hook teardown: scoring 10 e-commerce hooks, rewriting the weakest 5
A hook has one job: make the right person stop. To show what that means in practice, we wrote ten openings for ordinary products, scored each on three criteria and rewrote the five weakest. None of these hooks comes from a real brand; they’re built to show patterns you’ll find in your own ad account.
By the Dokiri Studio team · November 6, 2026 · 7 min read
In this article
The scorecard: clarity, relevance, novelty
Media buyer Bluesense Digital scores hooks in a fixed order: clarity, relevance, novelty, then specificity and credibility (video). We use the first three, from 1 to 5 each, because they decide who stops; the last two mostly decide who believes.
- Clarity. Would a stranger get what this is about in about three seconds, sound off? 1: no idea. 5: instantly.
- Relevance. Does it stop the people who would buy, and mostly them? 1: it stops anyone. 5: the buyer recognises their own situation.
- Novelty. Have they seen this opening a hundred times? 1: a stock template. 5: fresh for the category.
We judged each hook the way a muted viewer meets it: the first line, shown as text, over the first shot. Once an ad runs, the usual measure is hook rate, which practitioners compute as 3-second video plays divided by impressions (Meta’s definition of 3-second plays).
Ten hooks, scored
| Product | Hook: line over first shot | Clarity | Relevance | Novelty | Total /15 |
|---|---|---|---|---|---|
| Milk frother | “Wait for it…” over milk spinning in a glass | 1 | 1 | 1 | 3 |
| Non-stick pan | “Eggs slide off this pan with no oil.” A fried egg glides across a tilted pan | 5 | 4 | 2 | 11 |
| Clip-on book light | “POV: your partner sleeps while you read till 1 a.m.” A dark bedroom, one warm light on a page | 4 | 5 | 4 | 13 |
| Pet-hair roller | “The most satisfying thing you’ll see today.” Hair peeling off a sofa | 3 | 2 | 1 | 6 |
| Running belt | “Runners: stop holding your phone for 10 km.” A phone bouncing in a runner’s fist | 5 | 4 | 2 | 11 |
| Stain remover pen | “Nobody expected this at the Monday meeting.” Someone in an inflatable dinosaur costume dancing in an office | 1 | 1 | 5 | 7 |
| Garden kneeler | “Does gardening wreck your knees?” A gardener wincing as they stand up | 4 | 4 | 2 | 10 |
| Packing cubes | “Ten days of clothes. One carry-on.” A cabin case zipping shut over neat cubes | 5 | 5 | 3 | 13 |
| Kids’ drawing tablet | “Introducing our new drawing tablet.” The tablet spinning on a white background | 4 | 2 | 1 | 7 |
| Coffee canister | “You’ve been storing your coffee wrong.” A bag of beans closed with a clothes peg | 3 | 3 | 2 | 8 |
What the top scorers share. The book light and the packing cubes put a buyer’s situation on screen and make the product’s job obvious without naming it. Neither needs the sound on. The pan and the running belt are clear and relevant but familiar: they’ll work, and they’ll be the first to tire.
The running belt’s call-out. “Runners:” filters the audience, but by label. Bluesense Digital finds persona call-outs overrated: people dislike being labelled, and a narrow label caps how far an ad can spend (video). The bouncing phone already does the filtering.
Why weird-for-weird’s-sake hooks attract the wrong buyers
The dinosaur will probably have the best hook rate of the ten. That’s the trap. It stops people who enjoy costume videos, and nothing about a stain pen follows from it, so they leave when the product shows up, and the few who stay don’t remember what was sold.
Bluesense Digital’s argument goes further: a hook that catches everyone tells delivery to find more people who stop, not more people who buy. Organic content can live with that, since a million free views for a handful of sales costs nothing; a paid ad pays for every one of those views (video). That’s his reading of how delivery responds, not a documented Meta rule, but paying for the wrong views is a loss either way.
Surprise isn’t the problem; unrelated surprise is. Three questions sort a useful pattern interrupt from a gimmick:
- Is the odd image the problem or the product’s job, rather than a random spectacle?
- Would the buyer see their own situation in it within three seconds?
- Does the next line follow from it, or does the ad have to start over?
The five weakest, rewritten
Each rewrite keeps the product and changes what the first two seconds promise. The hook kind in brackets is the closest of Dokiri’s eight. Any number in a hook must be one you’ve measured: “20 seconds” only if your frother really does it in 20.
Milk frother (3 → 12). Before: “Wait for it…” over spinning milk. After [Result first]: café-style foam on a mug at a kitchen table, 7 a.m. “Café foam at home, in 20 seconds.” Pet-hair roller (6 → 13). Before: “The most satisfying thing you’ll see today.” After [Pain point]: black trousers covered in white dog hair, keys in hand at the door. “Dressed for dinner. Then the dog said hello.” Kids’ drawing tablet (7 → 12). Before: “Introducing our new drawing tablet.” After [Demo]: a child’s drawing vanishes at the press of a button. “One press, and it’s a blank page again.” Stain remover pen (7 → 12). Before: a dancing dinosaur at the Monday meeting. After [Pattern interrupt]: someone pours espresso down their own white shirt. “Yes, on purpose. Watch.” Coffee canister (8 → 12). Before: “You’ve been storing your coffee wrong.” After [Question]: a bag of beans held shut with a clothes peg. “Why does your coffee taste flat by week two?”
The pattern repeats: name a situation the buyer lives, show the product doing its job or the problem it ends, and keep the brand name for later. The stain-pen rewrite keeps the surprise, because the surprise is now the demonstration. It also has to be real: show only stains the pen actually removes.
The 8 hook kinds, mapped
Dokiri’s batch planner opens each concept with one of eight hook kinds. Here’s where our hooks land and what each kind needs to work.
| Hook kind | In this teardown | Works when | Watch out for |
|---|---|---|---|
| Question | Coffee canister (rewrite), garden kneeler | The viewer answers yes in their head | Questions that imply something about the viewer’s body or health |
| Pain point | Running belt, pet-hair roller (rewrite) | The pain is shown, not described | A pain the whole category already talks about |
| Result first | Milk frother (rewrite) | The result fits in one frame | Results the product can’t reliably deliver |
| Pattern interrupt | Dinosaur (bad), stain pen (rewrite) | The surprise is the problem or the demo | Spectacle unrelated to the product |
| POV | Book light | The scene is one the buyer lives | POV lines that describe nobody in particular |
| Bold claim | Packing cubes | Your product description backs it word for word | Anything you couldn’t prove on screen |
| Story | None here | The story is true: yours, or a real customer’s with permission | AI-drawn people presented as real customers |
| Demo | Non-stick pan, drawing tablet (rewrite) | The product’s job fits in two seconds | Showing anything the product can’t do |
Three of the original ten fit no kind at all: “Wait for it”, “most satisfying” and “Introducing”. That’s a useful test on its own: if you can’t say what kind of hook you wrote, it probably promises nothing. More templates in the hook library.
Score and test your own
- 1
Draft wide
The free ad hook generator writes eight openings of different kinds for your product, each with the first shot to show under it.
- 2
Score each one
Clarity, relevance, novelty, 1 to 5. Our cut-off: drop anything under 9, anything that comments on the viewer’s body, and anything you couldn’t show truthfully.
- 3
Check the bridge
Read each survivor followed by your ad’s second line. If the ad has to change subject, the hook belongs to a different ad.
- 4
Make three to five
Send a hook to the studio as your ad’s angle, or write the persona and angle in the Angle, offer or audience field and make a batch: each concept opens with a different hook kind, and ads.csv records each opening.
- 5
Read results in order
Hook rate first, then purchases from new people. A hook that wins attention but not sales is the dinosaur again.
Frequently asked questions
What makes a good e-commerce hook?
It shows the buyer a situation they recognise within about three seconds, with the sound off, and leads straight into the rest of the ad. Clear and relevant beats clever.
Are curiosity hooks bad?
No, but the curiosity has to be about the buyer’s problem. A gap that any scroller wants closed fills your results with people who were never going to buy.
How many hooks should I test per ad?
Practitioner advice ranges from three to ten per body. Three to five hooks of different kinds is a good start; wording tweaks of one hook teach you little.
Is the hook generator free?
Yes. You sign in with your e-mail the first time, with no password and no card, and each run gives eight openings, each with its first shot.
