Facebook offers efficient, effective split testing through its advertising platform. Here’s how to run split tests and (more importantly) why you should run them.
Split testing can be really useful for marketers. It can also waste time and money.
So let’s clarify the two essentials of split testing.
First, when formally split testing ads, you must have one – and only one – variable you want to test.
This means your ads will be identical except for the variable you’re testing.
For example, this business ran a split test with an ad that just used an image, and one that ran a video:
The ads are the same except that one has video content and one is an image.
It’s essential that you set up split tests based on only one variable, otherwise, when you get data on a winning ad, you won’t know which factor to credit.
Keep in mind that testing of ads with multiple variables happens by default at the campaign level on Facebook. When you set up a campaign with multiple ad sets, the algorithm will automatically channel budget to your highest performing ads. If all you need is to find the highest performing ad – but not necessarily know why it performs best – you can do that without split testing.
This re-enforces the purpose behind running a formal split test. In the split test, there is one variable you need to measure; it’s not just a matter finding the highest performing ad.
Which connects to the second essential of split testing.
When you split test, your goal is to gain data on a major ad variable. In other words, you’re not just testing if changing the color of your image text will make a difference.
Instead, you want to test on something that will inform your entire marketing strategy.
The example of testing whether video or images drive more results is this level of variable. This business discovered that while both ads drove a similar number of conversions (website visits), the video ads delivered more that 2x ROAS.
The results of this split test informed their larger marketing strategy. They started using more video to explain their services and make their pitch.
You can test the following variables through Facebook split tests:
- Optimization event
- Product set
Creatives are the actual design and text elements of your ads, and are probably the most commonly tested variables.
Again, to run a split test and get statistically significant data, you want to test high-level variables that will inform your marketing strategy, such as:
- Ads with images of people using products versus ads with just product images.
- Ads with distinctly different headlines.
- Ads with long versus short form text.
- Ads with distinctly different calls to action.
Change these variables but keep everything else identical so you know what influenced the results.
Audiences will help you understand what types of people are more likely to respond to your ads. For example:
- Different demographics like age or location.
- Custom audiences versus interest targeting.
- Different interest sets, such as people who follow a popular conservative news channel versus people who follow a liberal news channel.
Use audience insights to develop audiences you can split test based on interests.
E-commerce businesses can test product sets to learn which sets from their Facebook catalog perform better in those campaigns.
You could test variables like:
- Product sets with items less than $25 versus those with items more than $25
- Sets with the same types of products (like shoes) from different brands
- The same product sets with different offers, such as 50% off versus buy-one-get-one-free.
You can split test on placements to test if one platform works better for certain ads, such as:
- Facebook versus Instagram
- Automatic placements versus customized placements
- Mobile versus desktop
Budget and Set Up
The set up for Facebook split tests is fairly straightforward. There are a few different ways to approach it, but the system walks you through the process. Here are the basic instructions.
The system itself handles the delivery of your ads by dividing audiences into random, non-overlapping groups.
It is vital that you have an adequate budget so your ads get enough exposure to give the test statistical significance. Facebook will give you a suggested budget when you start as well as minimum budget.
You’ll set a time-frame to run your split test. It’s important that it’s long enough to gather data to determine a winning ad – and no longer. Facebook suggests a 4-day window to run your test. You can run for as short as one day or as long as 30. They also state that a winning ad can usually be discovered in 14 days or less.
Consider these time-frames and make sure the actions you want to measure can take place during them.
The nice thing about Facebook split tests is that it targets the thing you care most about. The winning ad set will be the one with the lowest cost per result.
When the test is complete, you’ll get an email with the results. You can go to the campaigns and view the details, which look like this:
In this case, one audience type engaged with the post for the lowest cost. Now the business can create more campaigns building off that audience, and analyze the audience to see how it reflects on their overall marketing strategy.
There’s a lot to learn and little to lose by using Facebook Ads split testing. If you’re not sure how to develop or target your ads, don’t guess.