After launching in beta earlier this year with PMD’s such as our partner Smartly.io, Facebook is now rolling out split testing for Facebook and Instagram Ads to self-serve users in Ads Manager and Power Editor. Split testing is a powerful feature that allows advertisers to truly see the effect that different variables have on their ads. You can test different placements, audiences, optimization strategies, and more to isolate and understand the impact of these variables.
Split testing on Facebook works by splitting your audiences into two random, mutually exclusive audiences. Each of these audiences is served the same set of ads with the exception of the one variable you’re testing. So for example, audience “A” may get ads optimized toward clicks while audience “B” may get ads optimized toward conversions. This would be particularly helpful if you’re close to the threshold number of 15 conversions per week that Facebook recommends when optimizing toward conversions. A few more important notes on split testing in Facebook Ads:
You can only test one variable at a time right now (hence split testing as opposed to multivariate testing)
The audiences are split based on people (Facebook User ID’s) and not cookies. This means the test will remain accurate across multiple devices per user since it is not based on cookies.
Facebook will monitor the results of your test and email a notification when a winner is declared
Split testing is rolling out gradually across ad accounts and will be available both in Ads Manager and Power Editor.
You must have a Business Manager account set up in order to run split tests
Facebook will give you a minimum budget for your split test based on the size of the audiences and estimated conversions.
Running split tests requires having an end date on your ad sets – the minimum duration is 3 days and the maximum is 14 days
Split testing currently supports the following 3 objectives
Mobile App Installs
You can test the following variables:
We’ve been running split tests since earlier this year and have found them to be a powerful method to scientifically determine which of our strategies are successful or not for our customers. In our experience, they work better with larger budgets and high conversion volume, but can work on smaller accounts as well if given enough time to run. Have you tested this feature out yet? Let us know in the comments below!