We ran a split test on these ad images and we were wrong 😭

Even when you think you know your brand and customers best, a good old split test comes along and schools you.

We are running a split test on a campaign for the Google Ads Guide using three different ad images:

• The first image is a nicely designed mockup of the guide on a laptop (image 1).

• The second is a graphic of the guide on a green background with a prompt to download (image 2).

• But we needed a third image so we zoomed into the pdf icon of the actual guide on the desktop and used that (image 3).

The ad was set live and we patiently waited until we had enough conversions to measure which image converted better.

Our expectations were that the nicely designed image would convert the most, the green one second and the pdf icon last or not at all.

What we actually found was that the nicely designed image (image 1) didn’t convert at all. Not one. That really surprised us.

The highest converting image was the zoomed in PDF icon (image 3), by a mile. The green background image (image 2) converted half of what the pdf icon achieved.

So what’s the moral of this story? You should always test multiple images and check the results.

In this case we tested three images with similar content (the cover of the google ads guide). But you can also choose to test between two very different subjects e.g. one could be the cover of the guide, and the other could be a photo of a frustrated person in front of a computer.

Whatever the outcome, it’s important to interpret and record the outcome. For the split test we conducted, we now know that customers respond to simple imagery and not overly designed images - this is something that’s being reflected in the other campaigns we’re running too. And this data now shapes how we design creative for our campaigns.

What if we didn’t split test images for this campaign? Imagine we thought that we knew best and wanted to only use the nicely designed laptop image. The outcome would be zero conversions, or very few conversions at a very high cost per lead. Are you launching campaigns using only one image? This could be the reason.