Advertiser method. Meta does not publish this as a rule. The mechanics below come from Meta’s documentation; the method itself is common advertiser practice, and results vary.
What should you test first on Meta ads?
Test the variable that answers your most important open question, one at a time. Meta lets you test creative, audience, placements or a custom mix, and recommends new testers start with one variable for the clearest insight. Write a measurable hypothesis first, keep the versions identical apart from that variable, and fix tracking before you start.
Test first whatever answers the question you most need settled, and change only that. This page fits when you have a running account, a handful of open questions and room for one clean experiment. It does not fit when tracking is unreliable or the budget cannot buy enough results; fix those first, because a test only reads what your measurement records. The Meta pages cited here document the variables, not a ranking of them.
Which variables does Meta let you test?
In an A/B test in Ads Manager, the kinds of variable Meta lists include:
- Creative: different images, text or creative types, managed at the ad level.
- Audience: different audiences or demographics, managed at the ad set level.
- Placements: for example automatic placements against chosen ones.
- Custom: several changes at once, to compare complex strategies on the same cost basis.
Meta says that if you are new to A/B testing, you should start with a single variable for the clearest insight into how it affects performance. Its own example hypothesis is about the performance goal: whether cost per result is lower when you optimize for landing page views rather than link clicks.
How do you pick between them?
The Meta pages cited here give no order, so pick by the question. The table matches common questions to a variable.
| Your open question | Variable to test | Where you change it |
|---|---|---|
| Which message or visual works for this audience? | Creative | Ad |
| Does a different group respond differently? | Audience | Ad set |
| Does a chosen set of placements differ from automatic ones? | Placements | Ad set |
| Is a different optimization goal cheaper per result? | Custom, on our reading: the pages cited here do not say which variable covers a performance goal | Campaign, ad set or ad, as Custom allows |
If two questions compete, you can start with the one you can change cleanly and read soonest. Creative and audience each have a dedicated page linked below.
What should be true before you test?
- The result metric is real. A test picks a winner on cost per result or a metric you choose, such as cost per purchase. If the purchase event is missing or double counted, the winner may be wrong.
- The audience is clear. Meta says the audience should be large enough, and should not be used in other campaigns running at the same time, because overlap can cause delivery problems and contaminate results.
- The budget can buy enough results. Meta says the budget should produce enough results to determine a winning strategy with confidence.
- The versions are identical except for the variable. Meta says this gives more conclusive results.
How do you write the hypothesis?
Meta’s example moves from a general question to a testable statement: from “do I get better results when I change my performance goal?” to “my cost per result will be lower when I optimize for landing page views.” You can copy the pattern: one variable, one metric, one direction. Write it down before you launch, so you read the result against it and not against whatever looks good afterwards.
What should you avoid?
Meta does not recommend informal testing by turning ad sets or campaigns on and off by hand. It says this can lead to inefficient delivery and unreliable results, and that A/B testing helps keep audiences evenly split. If you want to compare two strategies, use the A/B test tool and keep the test apart from the campaigns you scale.
FAQ
Can you test two things at once?
Meta offers a custom variable for several changes, but for new testers it recommends a single variable. If a test changes both the image and the audience, you cannot tell which change moved the result. Run a second test for the second question.
Should creative always come first?
The Meta pages cited here give no ranking, so no. Creative is one of the variables listed. Start with the question you most need answered and the variable you can isolate. If you have no such question, a creative test is a simple place to begin.
What if your budget is small?
Meta says a test budget should produce enough results to name a winner. Ads Manager shows an estimated power for the schedule and budget, and Meta suggests 80 percent or more. If the estimate is low, test a bigger difference or run longer.
Sources
- Selecting a variable for your A/B test (Meta, checked 3 Oct 2026)
- Best practices for A/B testing (Meta, checked 3 Oct 2026)
- About A/B testing (Meta, checked 3 Oct 2026)
- Create an A/B test by duplicating an ad set or ad (Meta, checked 3 Oct 2026)
- Tools to create A/B tests on Meta technologies (Meta, checked 3 Oct 2026)
Drafted with AI and checked claim by claim against the pages linked under Sources. Found a mistake? Tell us via the corrections policy.