The split test feature helps you to avoid overlapping of the target audience in two or more campaigns/adsets. This is especially useful for A/B testing.

Let’s take the example of an A/B test in which we want to determine the best creative between a video ad and a link ad using images. We would first need to have a structure like the one shown below:

Once you have the 2 identical campaigns created (ideally in pause), each of them using a different ad, go to your reporting view, select both campaigns and select the option Split Test in the action bar.

You’ll arrive to the screen where you can set up the details of the study, start and end dates as well as distribution of the population. By default the population will be split in equal parts, but you could modify this with the slider buttons. 

Once you click on Create, the study will be set and you could turn on your campaigns. The split test can’t be modified once it is created.

After running the test, please, follow this article to analyse the results.