How Long Should a Meta Ad Test Run?

Run a Meta A/B test for at least 7 days, and longer if your customers are slow to convert. That fits when you have one variable, a clear metric and a budget that can buy enough results. It does not fit if you want a quick read inside a day or two: Meta says tests shorter than 7 days may produce inconclusive results. Meta gives a floor and a ceiling, not a single right length.

What does Meta recommend for test length?

For the most reliable results, Meta recommends a minimum of 7 days. In Ads Manager you must schedule a test between 1 and 30 days, and Meta says A/B tests can run for a maximum of 30 days. In the Experiments tool, Meta says the test schedule is separate from the campaign schedule, and should cover the duration of your campaigns or enough time to yield reliable results.

You can use the 7 days as your starting point and change it only for a reason you can write down.

When do you need a longer test?

Meta says your ideal time frame may depend on your objective and business vertical. Its example: if your typical customer takes more than 7 days to convert after seeing an ad, run the test longer, such as 10 days, so the expected conversions have time to happen.

To use that, look at how long your own customers take between ad and purchase, and let the test run past that. Meta’s reason for the longer window is to allow enough time for the expected typical conversions to occur.

How do budget and power fit in?

Time is only half of it. Meta says a test should have a budget that produces enough results to determine a winning strategy with confidence.

When you set up a test, Ads Manager shows an estimated power: the likelihood of a statistically significant result, based on your schedule and budget. Meta suggests at least 80 percent. Meta says the calculation is only an estimate and does not guarantee a causal result. If the number is low, you can adjust the schedule or the budget. Meta’s setup page ties the estimate to those two, though its estimated-power page also mentions your objectives and other factors. The statistical significance page linked below explains how Meta reads the final result.

Should you check the test while it runs?

You can look, but be careful what you read into it. Meta says that while a test is still in progress, the cost per result will likely change. Once a test is running you cannot change the variable you chose. Changing the schedule or cancelling it affects the reliability of the results collected so far.

So set the end date up front, leave the variable alone and read the result at the end.

What if the test ends without a winner?

Meta says that if a test did not find a winner, you will see a recap and, where possible, suggestions. For example, if it did not run long enough to collect enough data, you will see a recommendation to extend it. You can edit the schedule from Experiments, but Meta says doing so affects the reliability of results collected so far.

Does the learning phase matter in a test?

Meta says ad sets are less stable and usually have a higher CPA during the learning phase, which usually ends after about 50 results in the week after the ad set’s last significant edit. You can check the Delivery column on both versions before reading a result: Meta says it reads Learning while an ad set is in that phase. That check is advice, not a Meta rule.

FAQ

Can a test run longer than 30 days?

Not as an A/B test in Ads Manager. Meta says A/B tests can run for a maximum of 30 days. If you need a longer comparison, run a follow-up test, or use a different tool; the holdout test page covers lift tests.

Can you stop early if one version is clearly ahead?

Meta notes that cost per result will likely change while a test is in progress. It also says cancelling or changing the schedule affects the reliability of results collected so far. Let it reach its end date unless you have a reason to stop.

Is 7 days always enough?

No. It is Meta’s recommended minimum, not a guarantee. Meta says to run longer if customers typically take more than 7 days to convert, and to check that the budget produces enough results. Estimated power is Meta’s own estimate of the likelihood of a statistically significant result from your schedule and budget.