18 Sep 2018
Facebook split-tests: 5 things you should know
So you’ve decided to split test your Facebook ads…
You’ve made a wise choice. Split-testing has proven time and time again to be the surest path to better digital marketing performance, no matter which channel a brand is using.
Just remember: split testing is only really worth doing if it’s done right.
Here are a few important things to keep in mind before you get your split-tests started.
1) The more elements you test, the better
Images, videos, copy, calls-to-action… there are many individual elements that make up every Facebook ad campaign. Somewhere within those elements lies the perfect, optimal combination to send your ad soaring into the performance stratosphere.
That’s why you’re split-testing.
The thing is, if you only split-test a small number of variants for your ad, the odds that you’ll stumble onto that perfect combination are actually pretty low. With so many elements at play, you could easily be looking at dozens (even hundreds) of intriguing versions of a single ad. While gut instinct and personal taste can sometimes help cut down this number, they often hurt more than they help.
Personal bias and tastes can be limiting. Just because you don’t think something will work doesn’t mean you are right, and just because you’ve never tried something before doesn’t mean you shouldn’t try it out now. Taking a few risks and split-testing a large number of variations of your ad’s visuals and copy is the best way to get personal bias out of the way and let the data speak for itself.
Sadly, Facebook’s current split-testing platform only allows you to test out 5 variations of your ad in each split test, which isn’t ideal. Luckily, there are ways around this limitation.
2) Variant quality is crucial
Split-testing low-quality images/videos, several variations of boring copy, or a bunch of weak calls to action won’t get you anywhere, no matter how many variations you try out.
Start your Facebook split-tests with high-quality images, copy, and calls to action. Doing so gives you the best chance of uncovering the ideal combination for your Facebook ad.
The Facebook timeline is a crowded place, and your ad will realistically have mere seconds to capture and hold your target audience’s attention. That’s why it’s worth investing the time and effort to produce eye-catching images and videos for each campaign you post. It’s also why ensuring that your ad’s copy is on-brand, engaging, and catered to the unique tastes of your target audience.
Remember that audience attention is only worth capturing if it results in those all-important clicks-through your campaigns have been missing. Your ads can’t make money if no-one clicks on them, after all! Our data has shown that optimized copy has a much bigger impact on click-through rates than optimized images and videos do.
3) Choose your performance metrics wisely
The more Facebook ad split-tests you run, the more performance data you’ll collect. But what to do with it all?
Before you start split-testing a campaign, it’s crucial that you set your goals for that ad first. Knowing what you really want to accomplish by running this ad will drastically impact how your split-tests are run. Looking for more video views, likes, and comments? Focus on your visuals and split-test the images you might include with this campaign. Looking for more clicks and conversions? Get your copy up to snuff and focus your efforts on testing out more engaging language.
Once you’ve decided what your goals are for a campaign, you’ll have a better idea of which metrics you should be focusing on to measure effectiveness in your split tests. Choose metrics which directly align with your goals, and aim your split-tests squarely on those metrics.
A campaign which results in a large number of “likes” is great, but if your goal was to drive traffic to your website and generate sales with your ad, those likes will be little consolation when your boss asks how it went.
4) Statistical significance matters
Effective split-testing is all about sample size.
If you don’t conduct your Facebook ad split-tests on a large enough audience, the performance data you end up with won’t be trustworthy.
Make sure that your sample size is big enough to be statistically significant. Obviously, budgetary concerns, total audience size, and the number of split-tests you’re running will come into play here. It’s important to be strategic when deciding how many people to include in your split-tests.
The more people you run tests on, the cleaner your performance data will be. It’s for you to decide which number fits the best for your brand/campaign. Choose wisely.
5) Trust the results
Fact: you might not agree that the ad variation that wins out in your split tests is the best of the bunch. You may not even like it at all. Here at Phrasee, where split-testing campaigns is a way of life, we’ve often encountered this phenomenon when running split-tests for our clients.
The thing is, data doesn’t lie.
If you’ve run your split tests on a statistically significant number of people and one variation of your ad performs better at achieving your goal than the other variations you tested, that’s the variation you should publish for your campaign.
If you decide to ignore the results of your split test and run the version that agrees with your personal bias and taste instead, you’ve just wasted a whole lot of time and effort running a split-test in the first place.
Again, data doesn’t lie. Trusting in the data and learning from it is what helps a brand’s Facebook ads get stronger and perform better over time.
If this all sounds like a whole lot of time and effort to invest, that’s because it is! The positive impact on your brand’s bottom line makes it worth it, though. And, if you’d like to save yourself the trouble and get the amazing benefits of ongoing split-testing anyway, our amazing AI Facebook ad optimization platform stands ready to give you a hand.