Guides · 17 October 2026
Testing UGC Ads on Meta: How to Compare Hook Variants in an A/B Test
Compare the hook variants of a UGC video in Meta Ads Manager: one variable, separate audiences, duration and the right metric.
Contents
To test UGC ads, you run the hook variants against each other in the A/B test in Ads Manager instead of running several campaigns side by side. According to Meta, the tool splits the audience into random, non-overlapping groups, so that the variants run under comparable conditions. One variable is tested at a time, and for hook tests that variable is the opening. As of 16 October 2026, checked against two Meta pages.
Why a single video is not a basis for testing
A single video gives you one number, but no comparison. Whether the cost per result is good or bad, you only know once a second variant has run under the same conditions. Anyone who orders just one video therefore has no basis for testing but a bet, as What Is UGC? puts it.
In my hook tests, it is three to four different openings for the same core video, not three to four videos with the same opening. Why the number of hook variants belongs in the briefing is covered in A UGC Briefing That Works.
One variable per test
According to Meta, a test should change only one variable at a time. As examples, the page names ad format, aspect ratio, video content (product or person), audio, placements and video length. On the other Meta page, ad creative, audience, placement and campaign objective are listed as possible areas.
For hook variants, this means the following, in my own application of the rule and not in Meta’s wording:
- The variable is the opening, meaning the first seconds.
- Everything after it stays the same: core video, format, ad text, audience, campaign objective.
- The variants should be as close in length as possible, because Meta itself lists video length as a separate test variable.
If you compare two completely different videos instead, you change the video content. You then learn which video performs better, but not whether the opening was the reason.
Setting up the A/B test in Ads Manager
According to Meta, the A/B test in Ads Manager compares several ad variants to find out which one delivers better business results. Up to five variants are possible. Several individual campaigns are no substitute for the test. According to Meta, the system does not split campaigns that run in parallel without the tool evenly. This distorts delivery and budget distribution and leads to inaccurate comparisons.
According to Meta’s page on advanced A/B tests, you create a new campaign, choose your ad objective and, at campaign level, activate the tool for A/B tests in Ads Manager. I do not describe any further click steps, because I could not back them up with any Meta page. What matters more is the preparation: each hook variant exists as a separate file, and an unambiguous name shows which opening is behind it. The ads run through the brand’s account. When a partnership ad fits instead is covered in Setting Up Partnership Ads.
Duration and budget
The two Meta pages give different durations. One mentions results within a few days and gives no fixed minimum duration. The other, Meta’s page on advanced A/B tests, gives a duration of at least two weeks or up to 30 days for the most reliable results. According to this page, first results are visible as soon as at least 100 events have been recorded for the cost per result. Meta advises, however, to wait until the end of the test before you evaluate.
I plan with the longer figure. On budget, Meta’s A/B test page gives no amounts, only one condition: it must be suitable so that the split is even and you can reliably determine the better strategy. I therefore cannot recommend an amount to you. It depends on your goal and your account.
Reading the result: which metric fits the goal
Meta names the cost per result as the measure for comparison. What counts as a “result” is determined by your campaign objective. In the case studies on Meta’s page on advanced A/B tests, the cost per lead and the cost per incremental purchase appear as metrics.
My recommendation follows from this: compare the hooks on the metric the campaign optimizes for. If it runs on leads, the cost per lead counts. If it runs on purchases, the cost per purchase counts. An intermediate metric such as the click-through rate can give hints, but does not replace the metric of the goal.
Two limits come with this. A winner holds for this test, with this audience and this product, not for all future ads. And Meta recommends testing regularly, so a single run is the beginning.
What the creator delivers for the next run
For a clean second run, these points belong in the briefing:
- Openings as separate files, each with the same core video behind it.
- One line per hook that describes what the opening does (for example question, problem or product in the frame). Without it, after the test you know which file won, but not why.
- New openings based on the winner, not a random new mix. That way the next test remains a comparison with one variable.
- Same length and same format, so that the opening remains the only change. More on formats in UGC Video Format for TikTok, Reels and Shorts.
How the collaboration works from booking to handover is described in Booking a UGC Creator.
- A UGC Briefing That Works: A Template to Copy
- Setting Up Partnership Ads: Running UGC Ads Through Instagram and Facebook
- UGC Video for Brands: Authentic Ad Videos Instead of Polished Spots
- Booking a UGC Creator: How the Collaboration Works
- What Is UGC? Definition, Formats, and What Brands Use It For
- Spark Ads With UGC: How the Ad Runs Through the Creator’s Account
Tell me briefly what you have in mind. I will get back to you with an honest assessment and a quote.
Send a requestFrequently asked questions
How many hook variants can I compare in a Meta A/B test?
According to Meta, up to five ad variants can be tested. With three to four openings for the same core video, that leaves room in the test.
Why not simply run several campaigns in parallel?
According to Meta, the system does not split campaigns that run in parallel without the test tool evenly. This distorts delivery and budget distribution and leads to inaccurate comparisons. The A/B test, by contrast, splits the audience into random, non-overlapping groups.
How long should a hook test run?
Meta's statements differ. One page mentions results within a few days and no fixed minimum duration, another gives a duration of at least two weeks or up to 30 days for the most reliable results. I plan with the longer figure and only evaluate at the end of the test.