Image ads are often more difficult to test than text ads as, with image ads, you’re rarely just testing ad 1 versus ad 2.
With image ads, you are often creating themes of images in various sizes and then seeing how each theme performs as a whole regardless of any one size.
For instance, you might have themes around:
The easiest way to test one theme versus another one in AdWords is to use labels. If you go into AdWords, its easy to first create labels and then apply them to your ads:
You can create a label for each theme and then apply the appropriate labels to your image ads so that you can easily sort, segment, and filter your images for each theme.
Once you have done this, testing theme by theme can be done in one of two ways:
You can wait until you think you have enough data to compare the themes against each other and download the ad data, use an excel pivot table (pivot off the label) and then compare the metrics of each label.
With Adalysis, you can setup tests by labels (or just sizes or label and size combinations).
Then the system will tell you when you have reached statistical significance so there’s no guesswork.
You can then see results automatically calculated by each label and even pause all the losers with a single click.
Regardless of which method you choose, you should start your image ad testing by first understanding how each theme does. Once you have a good idea of how to create your themes, then you can move to a more A/B testing approach for each ad group based upon how your are targeting your image ads across the display network.
Note: If you are only using placement targeting (so that your ads are always on the same placements); then you will want to A/B test ads by size by ad group since the ad display is consistent. If you are using keywords, topics, remarketing ,etc as your targeting option where the placements are not consistent, then theme testing with labels is the way to start.