A/B Testing: The Science of PPC Ad Management

If you've done any PPC management you have probably heard the phrase "A/B Testing." But what exactly does it mean? And how do you do it?

At Gnosis Arts, we pride ourselves on being data and experiment driven. But we also pride ourselves on being pragmatic. Therefore, we try always to simply complex concepts for our clients.

* What Is A/B Testing?
A/B testing refers to a scientific approach to PPC management. Basically how it works is, before launching a PPC ad campaign, you first create two campaigns. Both campaigns have the same parameters: the same keywords, the same ads, the same spend, and so on. In others words, you start by creating two exactly identical campaigns.

One of the campaigns will be your Control Group. If you've taken any basic level statistics courses, you might recall that a control group is the group in which the variables to be tested don't change. The second campaign will be the Test Group. This is the group where you test different variables, altering them here and there, to determine which one(s) is influential in the overall performance of the campaign.

A/B testing is useful because it enables you to isolate a single parameter in order to measure its efficacy. For example, in one campaign you can turn on the content network, leaving it off in the other campaign. Since the only difference between the two campaigns is the content network, any difference in performance can likely be attributed to that variable.

*Repeatability
According to the scientific method, to be truly scientific, any difference in campaigns would have to be repeated a significant number of times in order to be conclusive. This is to eliminate the possibilty that the effect of a particular variable did not happen by chance.

*What A/B Testing Accomplishes
With A/B testing, you have a better chance of maximizing the effectiveness of your ad campaigns. You can isolate variables such as Max CPC, test the strength of a particular managed placement, test the worth of a particular demographic factor, and more.

We recommend also keeping a campaign testing log where you record all you changes, on a day to day basis by date. This will help you visualize how each variable is affecting the campaign.