Advertising on Facebook requires many decisions to be taken, such as splitting your audience into targeting groups, and selecting a billing method or an optimization goal for your campaign. If you are an advanced Facebook marketer, the number of such decision parameters might increase even further. You might get overwhelmed by this complexity at first, but there is a scientific way to overcome all these challenges and gain a competitive advantage on Facebook: Split testing
Split testing on Facebook (also referred to as A/B testing) is a method of conducting controlled experiments based on comparing different ways of structuring campaigns. For Travel advertisers, split tests are undertaken to improve metrics such as: CTR, CPA, ROAS etc.
There are many split tests one can do on Facebook, and each gives the advertiser insights into how they can improve performance of their campaigns. Top four objectives that travel advertisers test when they start working with Adphorus are:
- Manual bidding vs. Marvin – to test our optimization algorithm, Marvin, against manual way of managing campaigns and seeing the added value of our automated bid & budget optimization engine
- Campaign / ad set structures – to test two or more possible audience targeting structures (including which audiences to target and how to split them) to find out which one works better in terms of efficiency and volume
- Ad image, messaging and C2A button – to test different creative types and chose between 2-3 ads with only a single change such as the image, messaging, or Call-to-Action button and test them against each other
- Pay per options – to decide whether to pay for impressions or link clicks in order to find the best performing pay per option for a specific campaign type
How to run a split test
Once a split test is decided, we are able to divide audiences into two or more groups randomly. Split audiences allow the advertiser to evenly test different options against each other and find out what works best. The aim of any split test is to get enough number of observations so that you can choose a winner with a certain confidence level. Therefore, all campaigns should be live until statistically significant results are achieved. Only after then, the split test can be stopped and the winner structure can be continued with. Our data science team will be involved in all steps of a split test, including setting up a fair test and presenting the results of the test scientifically.
In order to run a successful split test, it is crucial to keep these two points in mind:
- Have min 10 conversions per day – Budgets need to be high enough so that each ad set in the test campaigns can buy a minimum of 10 conversions per day. Thus, extra budget might be needed for a statistically reliable split test.
- Test one metric at a time – The most important aspect about a split test is that, one should only test one metric at a time. This means that when running a split test, all other metrics of the test campaigns need to be exactly the same. A split test can not evaluate more than one metric at a time.
Split test example: Testing the performance of different ad images
within the same target group
Maximize your Facebook Marketing performance via split testing
Split tests are essential stepping stones for any advertiser to maximize their Facebook marketing success. By trying different scenarios with regards to campaign structures, creatives and optimization, advertisers will be able to find out their optimal way of advertising on Facebook.
We are here to help with recommending which tests to take, analyzing the results and advising on the outcomes. Contact us and learn more about split testing and how to make use of it to boost your marketing efforts on Facebook.