Incipia blog

Facebook Ads Begins Rolling out Split Testing

Gabe Kwakyi | September 28, 2016

Finally. The ability to properly A/B test creatives in Facebook without its meddling algorithm abruptly choosing a winner!


Wrong, as of yet – but what Facebook is releasing is still very valuable.

Starting in mid-September, Facebook Ads began a gradual roll-out of its new split testing engine, picking website conversion as the first campaign objective to try this feature with.

Facebook's feature will enable advertisers to specify a particular variable to split test, and then show randomized, mutually exclusive and non-overlapping segments of either a targeted audience or duplicated ads in the ad set running the split test, with the winner sent via email. Advertisers will be able to select one of two variables for up to 3 splits (A/B/C), including:

  • Saved audiences (e.g. lookalike 1% vs lookalike 2%)
  • Bidding optimization (e.g. optimized for link clicks vs impressions)
Unfortunately, it doesn't appear that creative testing is on Facebook's radar just yet, and split tests won't come cheap: Facebook recommends a minimum budget of $3,000, with a minimum test period of 3 days and a maximum of 14 days.

While Facebook split testing tools have already been available to large customers spending hundreds of thousands of dollars with Facebook, for the majority of advertisers split testing has been an extremely frustrating task. Creative split testing options at current include:
  1. Using a creative testing service like Adspresso – this costs money and doesn't provide true split testing.
  2. Manually turning ads off and on – this method runs the risk that the results will be contaminated because Facebook's algorithm may choose to show the ads to different subsets of the audience and also because different days of the week often perform differently. This method also requires manual calculation of significance levels.
  3. Testing different creatives in different ad sets targeting the same audience – this can also contaminate the results as the two ad sets and ads within them will compete with one another and lead to favoring, because only one ad from an account is eligible to serve to one person. This method also requires manual calculation of significance levels.
The Bottom Line: Facebook split testing now enables select advertisers to better optimize their ad campaigns, starting with audiences and bidding optimization. While limited in scope, Facebook's split testing has promise to significantly improve results and ROAS for advertisers.

Original Source: Ayo Oluwole via Facebook.