Guide · 8 min read
How to test Facebook ad hooks
By Pavan, Founder, FirstReaction · Updated
Short answer
Test Facebook ad hooks by changing only the opening (the first one to three seconds of a video, the first image, or the first line of copy) while keeping the rest of the ad the same. Make three to five different types of hook, compare hook rate (3-second plays divided by impressions) first, then hold rate and cost per result.
What counts as a hook
The hook is whatever has to work in the first moment, before someone decides to keep scrolling:
- Video: the first one to three seconds, including the first frame, the first words spoken and any text on screen.
- Image: the first thing the eye lands on: usually the product, a face, or the biggest line of text.
- Copy: the first line of the primary text, which is all most people see before "…more".
- Carousel: the first card. Every later card only gets seen if this one works.
A hook test changes only that opening and keeps everything else the same. Otherwise you can't tell whether the hook or something else made the difference.
The numbers to watch
| Metric | How to work it out | What it tells you |
|---|---|---|
| Hook rate | 3-second video plays ÷ impressions | How many people stopped |
| Hold rate | ThruPlays ÷ 3-second video plays | How many of those kept watching |
| ThruPlay | Meta's count of plays to the end, or at least 15 seconds | How many watched most of a video |
| Click-through rate | Link clicks ÷ impressions | Whether the whole ad made people act |
| Cost per result | Spend ÷ your goal event | Whether it was worth it |
Hook rate and hold rate aren't built-in columns in Ads Manager; create them as custom metrics from 3-second video plays, ThruPlays and impressions. Definitions vary between teams (some use 15-second plays for hold rate), so pick one and keep it.
What's a good hook rate? It depends heavily on how warm the audience is. As a rough guide, the ranges FirstReaction calibrates its predictions to put a cold audience at about 10–22% and a retargeting audience at about 32–48%, before adjusting for industry.
Nine hook types worth testing
- The problem: "Still waking up with dry skin after cleansing?"
- The result first: open on the after, then show how.
- A bold, checkable claim: one specific number or fact you can back up.
- A question the viewer answers in their head.
- Social proof first: a review, a rating, a press mention.
- A pattern interrupt: an unexpected movement, angle or sound in the first frame.
- A creator talking to camera, mid-sentence, as if continuing a conversation.
- The product in action, close up, in the first second.
- The offer, when the audience already knows you: price, discount, or what's included.
Make your three to five test hooks different types, not five wordings of the same idea. You learn more from "problem vs result vs social proof" than from five versions of one line.
How to test hooks live
- Write or cut 3–5 openings for the same ad. Keep the body, end card, copy and landing page identical.
- Put them in one test: a Meta A/B test for a clean answer, or several ads in one ad set when you only need a direction.
- Give each enough impressions before judging. A few hundred impressions per hook is too few to trust.
- Read hook rate first, then hold rate, then click-through and cost per result. A hook that stops people but attracts the wrong ones shows up as a high hook rate with a weak cost per result.
- Keep the winning hook and test the next part: the middle, the offer, or the end card.
How to pre-test hooks before spending
You can filter hooks before any of them go live:
- Video hooks: upload each version (same video, different opening) to an AI audience simulation against the same audience. Compare the scroll-stop score and the video's hook score for each, and read which buyer types each opening loses.
- Copy hooks: run a bulk test with several first lines and headlines against one creative, and let it rank the combinations.
- Image hooks: check what people are predicted to notice first. If it isn't the product or the offer, the hook is working against you.
Then take only the strongest one or two into the live test.
Example: when the hook isn't the problem
We simulated a live creator video ad for Sahajan's cleansing oil ($41) against a cold US audience of women 22–40 who wear makeup and SPF and read reviews before buying skincare. Each version was run three times.
| Version | What changed | Likelihood to act | Predicted hook rate | Video hook score |
|---|---|---|---|---|
| Original (46 seconds) | — | 58.7 | 18.5% | 83 |
| Same video, new copy | Headline and primary text only | 56.2 | 19.1% | 87 |
| Re-cut (24 seconds) | Shorter, before-and-after, ingredients on screen, price end card | 57.7 | 19.3% | 85 |
| Re-cut v2 (27.6 seconds) | v1 plus a real customer review card and the rating | 66.8 | 19.7% | 90 |
The original's hook was already strong: an engaging creator, on-screen text, and a clear demonstration in the first seconds. Changing the opening barely moved the predicted hook rate. What moved likelihood to act was what came after the hook: social proof and the price, in the re-cut that added a real review.
The lesson: check your hook rate before rewriting the hook. If people already stop, a new opening won't fix a weak middle or a missing reason to buy.
Simulated on FirstReaction in October 2026. Scores are predictions from a simulated audience, not real campaign results. Not affiliated with Sahajan.
To pre-test your hooks before you spend, you can start free on FirstReaction with 25 credits.
Frequently asked questions
What is a good hook rate for Facebook ads?
It depends on the audience. Cold audiences stop far less often than past visitors or buyers. As a rough guide, FirstReaction calibrates cold audiences to about 10–22% and retargeting to about 32–48%, before adjusting for industry. Compare your hooks with each other, not with a universal number.
How many hooks should I test at once?
Three to five, each a different type (problem, result, social proof, offer). More versions split a small budget too thinly to read.
How long should a hook test run?
Until each hook has enough impressions to compare, usually several days. A few hundred impressions per hook isn't enough to trust a difference of a point or two.
Do hooks matter for image ads?
Yes. For an image, the hook is the first thing the eye lands on: the product, a face or the biggest line of text. For copy, it's the first line before "…more".
Can I test hooks before spending money?
Yes. Upload versions with different openings to an AI audience simulation against the same audience and compare their scroll-stop and hook scores, then take only the best one or two into a live test.
Sources
Related guides
- How to test Meta ad creatives with AIWhat AI can read in a Facebook or Instagram ad, four ways to test creatives with AI, a step-by-step method, a prompt for any chatbot, and a real before-and-after holiday ad.
- How to test Facebook ads before launching themA practical 5-step workflow to test Facebook and Instagram ads before you spend: research, creative QA, AI audience simulation, a small live test, then scale. With a real before-and-after example.
- How to validate ad creatives on a small budgetHow to test Facebook and Instagram creatives when you can't afford to test everything live: free checks, a run/fix/kill framework, a $10–$20 a day plan, and the real numbers from our own small campaign.