What a split test is
Checked 29 October 2026.
A split test, which Meta calls an A/B test, compares two versions of your ads under fair conditions. Meta's help page on A/B testing puts it this way: "A/B testing lets you compare two versions of an ad strategy by changing variables such as ad images, ad text, audience or placement. We show each version to a segment of your audience and ensure nobody sees both, then determine which version performs best."
The key part is "nobody sees both". Each version reaches its own share of the audience, so the two versions are not competing for the same people.
Why not just run one ad, switch it off and run the other? Meta says that approach "can lead to inefficient ad delivery and unreliable test results." The two ads run at different times, to different people, under different conditions, so you cannot tell whether the ad or the timing made the difference.
What you can test
Meta's page on test variables lists four kinds:
- Creative: images, video and ad text.
- Audience: who sees the ads.
- Placement: where the ads appear.
- Custom: another variable you define.
Whatever you test, the rule is the same. Meta's best practices say "Test only one variable for more conclusive results", and its variables page adds: "If you're new to A/B testing, we recommend starting with a single variable".
The reason is simple. Meta's page on testing more than one variable says: "When you're testing more than one variable, you won't know exactly which variable led to the winning ad set." If version B has a new headline and a new image, and it wins, you still do not know which change did the work. Change one thing, and the winner tells you something you can use again. That is also the idea behind what a microtest is.
How to set one up
Checked 29 October 2026.
Meta's help page on creating an A/B test describes two routes.
The Ads Manager toolbar. "Generally, we recommend creating an A/B test with the Meta Ads Manager toolbar, which uses an existing ad campaign or ad set as a template for your test." This is the quickest route when you already have a campaign and want to test one change against it.
The Experiments tool. Meta suggests it "when you want the flexibility of fine-tuning your ad campaigns before starting your test, or when you want to use several existing ad campaigns".
Whichever route you take, write a hypothesis before you start. Meta's own example: "My cost per result will be lower when I optimize for landing page views." A hypothesis names the one change and the result you expect, which keeps the test focused and makes the outcome easy to report.
How long and how much to spend
Checked 29 October 2026.
On length, Meta's best practices are clear: "For the most reliable results, we recommend a minimum of 7-day tests. A/B tests can only be run for a maximum of 30 days, but tests shorter than 7 days may produce inconclusive results." The schedule you set must be between 1 and 30 days. If your customers usually take more than 7 days to convert, Meta's example suggests running the test for about 10 days.
On budget, Meta says: "We recommend using the same budget for both versions in a test to ensure a fair comparison." Uneven budgets give one version more chances to win.
On audience, Meta advises an audience large enough for the test, and one you are not using in another campaign at the same time. Otherwise your other campaign competes with the test for the same people.
How Meta picks a winner
Checked 29 October 2026.
Meta's help page on how winners are determined says the winner is picked on cost per result, or on cost per incremental purchase when you use a holdout. Meta "simulates possible outcomes tens of thousands of times" and "statistically calculates a winner with a certain confidence percentage".
Sometimes there is no clear winner. If too few results came in, Meta tells you: "There weren't enough results to calculate an accurate winner." When the test ends, you get a notification in Ads Manager and an email.
A test with no clear winner is a result too. It tells you the change you tested did not matter much, which is worth knowing before you spend more on it. Report it that way, as a tie, not as a failure and not as a win. For a way to write that up, see a report template with a test record, or how to report test results to clients.
The creative test option
Checked 29 October 2026.
Meta also offers a separate creative test. It lets you "compare up to 7 creative variants", so 2 to 7, inside one campaign. It uses the Highest volume bid strategy and no more than 20% of the budget. Meta says it "does not make any automatic changes", and "A confidence level is not included." You see results by email, through the beaker icon or in Experiments.
So the creative test suits comparing finished creatives inside a live campaign. An A/B test suits a question where you want a winner with a confidence percentage.
Common mistakes
- Changing several things at once. The winner cannot tell you which change worked.
- Switching ads on and off instead of running a real test. Meta says this can give unreliable results.
- Stopping before 7 days. Meta says shorter tests may be inconclusive.
- Uneven budgets. Meta recommends the same budget for both versions.
- Reusing the test audience in another campaign. The two compete for the same people.
- Calling an unfinished test a win. Until it finishes, being ahead is not a result.
When a smaller message test fits better
1Signal is our product. It is not a replacement for Meta's A/B test. It answers a different, earlier question: which message should your ads say?
1Signal "runs your lines as small real Meta ads and shows you which one people click", in five focused tests, "each changing one thing: pain, headline, offer, image and ad text." The winner of each carries into the next. You see "the exact line, with its click rate, clicks and spend", and "Only completed test results count." "Nothing launches or spends without your approval."
Plans start at $29 a month, and Meta ad spend is separate. To be plain about what you get: the result is a click rate on Meta, not sales. It tells you which line people responded to in the ad.
Get your report and first-test draft free.
When Meta's own test fits better
Use Meta's A/B test for audience, placement, or budget and bidding questions, or when you want a winner judged on cost per result or a holdout. Use Meta's creative test to compare 2 to 7 finished creatives inside a live campaign. 1Signal does not test audiences, placements or budgets. For tools built around creative testing, see creative testing tools for agencies, and compare 1Signal plans if a message test fits.
Meta A/B test vs Meta creative test vs 1Signal microtest
| Meta A/B test | Meta creative test | 1Signal microtest | |
|---|---|---|---|
| What it compares | Versions of a campaign or ad set: creative, audience, placement or a custom variable | Creative variants inside one campaign | Lines that change one thing: pain, headline, offer, image or ad text |
| How many versions | Two versions of an ad strategy | 2 to 7 creatives | Five focused tests, each changing one thing |
| How it picks a winner | Cost per result, or cost per incremental purchase with a holdout, with a confidence percentage | No confidence level included | The line with the best click rate in a completed test; ties are labelled |
| Budget | Meta recommends the same budget for both versions | No more than 20% of the budget | Plans from $29 a month; Meta ad spend is separate |
| Best for | Audience, placement, budget and bidding questions | Finished creatives in a live campaign | Which message gets clicks before you build a full campaign |
Meta facts from its help pages, checked 29 October 2026. 1Signal is not affiliated with Meta.
FAQ
How do I split test Facebook ads?
Create an A/B test in Meta Ads Manager, either from the Ads Manager toolbar using an existing campaign or ad set, or in the Experiments tool. Change one variable, give both versions the same budget, run it for at least 7 days, and read the winner Meta reports on cost per result.
What can I A/B test in Meta Ads Manager?
Meta's help pages list creative, such as images, video and ad text, plus audience, placement and a custom variable. Meta recommends testing one variable at a time, because when you test more than one, you will not know exactly which change led to the winning ad set.
How long should a Facebook A/B test run?
Meta recommends at least 7 days for the most reliable results, and a test can run for at most 30 days. Tests shorter than 7 days may be inconclusive. If your customers usually take more than 7 days to convert, Meta's example suggests running the test for about 10 days.
How does Meta decide the winner of an A/B test?
Meta compares cost per result, or cost per incremental purchase when you use a holdout. It simulates possible outcomes tens of thousands of times and reports a winner with a confidence percentage. If there are too few results, it says so. You get a notification in Ads Manager and an email.
What is the difference between an A/B test and a creative test?
An A/B test compares versions of a campaign or ad set, such as two audiences, and reports a winner with a confidence percentage. Meta's creative test compares 2 to 7 creatives inside one campaign, uses no more than 20% of the budget, makes no automatic changes and includes no confidence level.
When is a smaller message test a better fit?
When you want to know which pain point, headline, offer, image or ad text gets clicks before you build a full campaign. 1Signal is our product: it runs small real Meta ads that change one thing at a time and shows the winning line with its click rate, clicks and spend.