I’m always surprised when split-testing online ads. Lately, I really can’t describe why one ad out-performs the others.
I set up an acquisition campaign on Facebook to drive more followers to our company page. We set up 6 ads utilizing the normal optimization techniques. The goal is to test all six and when the top performer emerges, we will adjust our budget to the best ad.
What I found interesting is that all six were fairly similar with slight changes to test for effectiveness. However, one ad in particular wasn’t visually different but was significantly different in the execution of the design. The other ads used tested techniques of showing faces along with our company logo, but one ad did not have a graphic of our logo placed on the image, but instead was a picture of one of our news trucks with our logo on the side. This ad is the highest-performing ad.
Now I don’t know for sure, but I theorize it has something to do with Facebook’s recognition technology. I believe that Facebook cannot read the actual text of the graphic because it’s embedded within the picture, instead of placed on a layer inside Photoshop.
I can’t really prove this, but I find it peculiar. Therefore, after this ad campaign has ended, I’m going to attempt the same scenario with ads that are compositionally different. I will update what I find out.