The 3-second hook: what actually stops the scroll
Hooks fail in the same five ways across every platform. Here's a working taxonomy of what stops a scroll, what doesn't, and how to test which one your ad needs.
Every paid creative lives or dies in the first 3 seconds. Not 5. Not 10. Three.
By second 3, a viewer on Reels, TikTok, or Stories has already decided whether your ad is worth more attention than the next swipe. The decision isn't conscious. It's the same instinct that lets you skip a flyer at a coffee shop without reading the headline. The brain runs a fast scan, finds nothing it cares about, and moves on.
You can't talk a viewer out of that decision. You can only design around it. Most ads fail this part because their hook was treated as a creative flourish instead of a load-bearing structural choice.
Here's what actually works, what doesn't, and how to test which kind of hook your ad needs.
What a hook is doing, mechanically
A hook isn't a sentence at the start of the ad. It's the bundle of visual, auditory, and textual signals that fire in the first 3 seconds.
In that window the viewer is asking exactly one question, even if they couldn't articulate it: is this for me?
A working hook answers that question fast enough that the viewer holds for a fourth second. Everything else in the ad, from the offer to the CTA, only exists if the hook bought you the time to deliver it. This is why the Hook dimension is scored first in our framework. Nothing downstream matters if the hook doesn't earn the attention.
The five hooks that work
Across the creatives we've scored, five hook archetypes show up disproportionately in the top quartile.
1. Named pain
You say the exact problem your audience is having, in their own words, in the first 3 seconds.
"If your ad creative tests are taking 6 hours per week, you're spending more time deciding than running them."
The audience hears their own week described back to them. The pattern interrupt isn't visual; it's recognition. They stop because the ad recognized them, not because they recognized the ad.
This is the highest-leverage hook for problem-aware audiences. It fails on cold traffic, who don't yet identify with the pain.
2. Pattern interrupt
A visual or audio break from the normal cadence of the feed. The feed runs on motion conventions: stable shot, slow zoom, person talking to camera. A hook that violates the convention in second 1 buys attention.
Examples that consistently score high on Hook:
- A flash cut from black to a high-energy moment in frame 1
- A face entering from an unexpected angle
- A product used wrong in a way the viewer notices before they understand it
The trap: pattern interrupts wear out fast. By the time you've seen six ads open with the same fake-out, the interrupt is the new pattern. Rotate.
3. Specific contrast
You show two states side by side with no preamble. Before and after. Cheap and expensive. Slow and fast. The viewer's brain finishes the comparison automatically and they want to know which side they're on.
This works because contrast forces a question without asking one. The viewer is already inside the ad, mentally, by second 2.
The discipline: the contrast has to be specific. "Bad vs good" is too vague. "$340 in ad spend before vs $340 in ad spend after" lands.
4. Numbered claim
A specific number in the first 3 seconds anchors the rest of the ad against a measurable promise.
"We scored 2,847 ads in the last 90 days. The same six mistakes showed up in 71% of them."
Numbers buy credibility because they're falsifiable. Round numbers feel made up; specific numbers feel measured. "Around 70%" scrolls; "71%" stops.
The trap: the number has to be doing work in the rest of the ad. A specific number with no payoff feels like a bait.
5. Direct address
You name the role or situation of the viewer directly in the first 3 seconds.
"Solo marketers running 3 clients: this is the workflow that saved me 8 hours last week."
The viewer has to opt out instead of opt in. They have to decide "no, this isn't for me" before they can scroll, and that micro-decision is enough to buy a fourth second. By second 5 you've earned the right to deliver the offer.
This works best when the audience is narrow. Direct address to "marketers" is too broad to feel personal. Direct address to "solo marketers running 3 clients" feels like you wrote the ad for one person.
The four hooks that consistently fail
The other side of the data is just as useful. Four hook patterns show up in the bottom quartile across categories.
Product-first. The ad opens with a shot of the product. The viewer can't tell what it does, who it's for, or why it's relevant. They scroll. This pattern accounts for the largest single chunk of low Hook scores in our dataset.
Brand-first. The ad opens with the logo, the tagline, or a brand identity moment. Brand-first works for awareness campaigns at scale; it fails on cold performance traffic where the viewer doesn't yet know or care about the brand.
Generic benefit. "Save time on your work." "Get more done." "Take control of your business." Anything that could be the opening of any ad in any category. The viewer doesn't recognize themselves, so they don't stop.
Slow build. A 5-second establishing shot of a person walking into a frame, sitting down, opening a laptop. By second 3 nothing has happened. The viewer is gone before the hook even loads.
If your ad opens with any of these and you're wondering why your CPM is fine but your hold rate is awful, you've found the regression.
How to test which hook your ad needs
Hooks are not interchangeable. The same product will need a named-pain hook for problem-aware traffic and a numbered-claim hook for cold traffic. Picking the wrong category is the most common reason "good" hooks fail.
Use this short test before you ship:
- Where is the audience on the awareness ladder? Unaware, problem-aware, solution-aware, product-aware, brand-aware? The further left, the more your hook needs to do work to surface the problem before the offer.
- Has the viewer felt this pain in the last 7 days? If yes, named pain wins. If no, pattern interrupt or numbered claim has to do the work named pain would have.
- Are you the cheaper alternative to a known reference? If yes, specific contrast hooks tend to outperform.
- Is the audience narrow enough to name? If yes, direct address compounds. If no, narrow first, then write the hook.
The pre-flight test takes 90 seconds. Most teams skip it. That's why most teams ship hooks that match the founder's gut rather than the buyer's stage.
What "good" looks like in the framework
When we score a creative, Hook is the first dimension because nothing downstream survives a failed hook. A 9 on Hook with a 6 on Offer Alignment will outperform a 6 on Hook with a 9 on Offer Alignment, because the second creative never earned the right to deliver the offer.
A 9 on Hook usually has three things:
- A specific signal (named pain, specific number, named role) in seconds 0-3
- Visual or audio motion that breaks the feed's cadence
- A payoff that arrives by second 6, not second 15
A 4 on Hook usually has one thing: a product or brand opener with no signal of who the ad is for or why it's relevant.
You don't need a 10. You need a 7-plus consistently. The teams that ship 7-plus hooks regularly do it by treating hook as a structural choice, not a creative decoration.
The shift to make on your next ad
Before you write the next ad, write the hook archetype on a separate line and pick from the five that work. Don't write the opening line first; pick the archetype, then write into it.
If you can't pick an archetype, you haven't done the work to know who the ad is for. Step back. Write the audience. Then write the hook.
Want to see how your current ad scores on Hook (and the other 7 dimensions)? Score it on Adverdly free. You'll get the score, the rationale for why the hook works or doesn't, and the priority fix at the top.
Then write the next one with the archetype in mind. Hold rate will move in a week.
