Create Facebook Split Test with Ad Set Duplication

Want a quick way to split test different audiences with your Facebook ads? A new feature will help you.

Note that this is different from Facebook’s primary built-in split testing feature. With that feature, you would set up a split test with a new campaign based on creative, delivery optimization, audience, or placement.

Facebook Ads Split Test

This new split testing option is related only to actively running ad sets. Let’s take a closer look…

When You’d Use This

Unlike the primary built-in split testing feature where you’d plan a test from the start, this new option is for an actively running ad set.

In other words, an ad set is running and you decide at that point that you want to try something else. For example, the ad set is running fine and you don’t want to edit or stop it. But you’re curious how a variation of the targeting would perform.

Create a Split Test by Duplicating

Select the ad set that you want to split test and click the “Duplicate” button.

Facebook Ad Set Duplicate

You should then see an option to create a new split test.

Facebook Ad Set Split Test Duplication

The only options for me are Age and Gender or Saved Audience. Until recently I only had Age and Gender.

Note that within the Help Center, Facebook indicates that you can create a split test by duplicating an ad set using the following variables:

  • Audiences
  • Delivery Optimizations
  • Placements
  • Product Sets
  • Creative

That would more closely align this method with the built-in split testing feature. You may have these options. I don’t at the moment.

After making your selection, click the “Continue to Test and Learn” button.

Facebook Ad Set Split Test

First, determine how your targeting will be different in Version B.

Facebook Ad Set Split Test

You will have the ability to edit the ad set name, schedule, and budget.

Facebook Ad Set Split Test

Facebook will provide an “Estimated Test Power,” which is the likelihood to detect a difference with your test.

Facebook Ad Set Split Test

You may want to increase budget to improve the test power. When you’re ready, click “Create Test.”

Facebook will then generate two new campaigns, one for each version…

Facebook Ad Set Split Test Duplication

Split Test Results

You’ll view the results within Ads Manager as you normally would…

Facebook Ads Split Test

But you can also view the test details within the Test and Learn section. First, you can access Test and Learn within the main Ads Manager menu…

Facebook Ads Split Test

There, you will see your test results under “Learn.”

Facebook Ads Split Test

In my case, Facebook found a winning campaign before the test completed. Click “View Report” for details.

Facebook Ads Split Test

While the results of the two campaigns were very similar, Facebook found a winner anyway. The test reveals that there would be a 66% chance of getting the same results if tested again.

Facebook Ads Split Test

Things to Consider

First, know that these two new campaigns that are created will run in addition to your active ad set. You may want to pause your original ad set in the meantime to prevent unnecessary overlap.

Second, the budget will default to half of what was remaining in the existing ad set’s budget. As mentioned earlier, you can edit that during the duplication process.

Always consider volume required to find a winner. If you are split testing a conversion campaign, how much do you need to spend to yield an adequate sample size? The more the product costs, the more you should expect to spend to get those sales. Facebook is less likely to find a winner if you aren’t generating an adequate sample size.

You can increase volume either by extending the length of the campaign or increasing the daily budget.

Split Test by Editing an Ad Set

You can also start a split test by editing an ad set. While this opens up more split testing flexibility than the duplication option, many restrictions apply.

We’ll cover this later in a separate post.

Your Turn

Have you experimented with this split testing feature? What kind of results are you seeing?

Let me know in the comments below!