What does A/B testing mean in marketing?

On paid social, the cleanest tests change one thing: the angle, the format or the hook. Changing colours or button text between two otherwise identical ads rarely tells you much.

A batch is a test of formats on one angle: 5 videos in 5 different formats and 5 statics in one ad set. Across batches, each new angle is a test of the reason to buy.

Our free A/B test calculator and sample size calculator show whether a difference is likely to be real.

When a claymation version of an angle wins and the talking head of the same angle loses, the next batch builds on the format. When every format of an angle loses, the angle is the problem.

We read the results and build the next batches from the angles and formats that won. More on how is on our performance marketing page.

Comparing two ads inside one Meta ad set is not a randomized test: Meta sends each ad to the people it expects to respond.

Decide the sample size before you start, and do not stop the moment the result turns significant. A result that is significant at 95 percent does not mean B is 95 percent likely to be better, and it does not say how big the real difference is.

On landing pages, test one change at a time, such as the headline or the first intake question.

Statics are fast to read, cheap to change and easy to test. Motion graphics suit testing many headlines on the same visual.

Every test ad still goes through review: Meta says its ad review system is designed to review all ads before they go live. Who runs the tests is covered on our media buying page. Source: Meta: About the enforcement of Meta Advertising Standards, checked on 3 November 2026.