Guide How to do an A/B Testing (With Excel)

One of the most impactful optimizations that you can do in a campaign is A/B Testing. This test allows advertisers to compare results of two different ads in the same campaign/budget.

This example available here can be used on display/search/mobile/native/video campaigns. You just need to adapt the metrics that you use as KPIs.

In past I used this model for Adwords (GDN and Search campaigns), in Native (like Taboola and Outbrain), and in Programmatic (all the DSP’s like Appnexus and Doubleclick).

We start with the date. This will allow us to identify when the test began. Then we have the ad variation, so we can take conclusions about the variations. If we do many tests with the same variation, maybe we can conclude after that a common element can be used as indicator for all the creatives. Like a banner with a button or without button. The third dimension is the Campaign, so you can pause or reduce the weight of the ad, the ID of the ad if that helps, and then the metrics:

  • Impressions (how many impressions your ad had. This number should be similar in all ad variations on the test)
  • Clicks (this are the clicks your ad had)
  • Spend (This number should be the same in all ad variations, like the impressions)
  • Conversions (If you have a performance campaign where you use conversion tracking)
  • CTR (Click Through Rate – This is related with the clicks on the total number of impressions.
  • CVR (Conversion Rate – This is related with the number of conversions on the clicks that the ad had)
  • I2C (Impressions to Conversions – On this dimension you can see how many impressions and Ad variation needed to have a conversion)
  • CPA (This metric is normally the last one and the most important. The Cost per Action/Conversion)

Leave a comment if you have any doubts how to use the metrics and the file.