Yesterday, we wrote about the importance of A/B testing your ads. A/B testing your ads is one of the single best things you can do to increase your ad revenue; it’s quick, easy, and in many cases, free. But how do you do it? How can you test whether one ad unit or another is working well for you? How can you test whether display ads or text ads are better for your website? Every website is different, and you can’t just assume anything when it comes to ads on your website.
Fortunately, there are several free tools you can use to run A/B tests on your ad units. The most fully-developed of these is Google Website Optimizer, so that’s what we recommend.
Although you’ll have to manually update the changes to the code you want to A/B test each time you run a different experiment, it’s worth it to see just how your ads are performing.
The answers may surprise you.
Steps to Run A/B Test
Running an A/B test using Google Website Optimizer isn’t that challenging, but before you do so, you need to do some preliminary work.
First, like any good experiment designer, you need to develop a hypothesis. What are you trying to test? An example of a good hypothesis would be “an ad unit on the right-hand side of the website will perform better than an ad on the left-hand side of the website”. Some publishers might just take that assumption and run with it.
Once you have your hypothesis, take a moment to write out exactly what you’ll be testing. In any other type of experiment design, this would be considered defining your control group and experimental group. In this scenario, it’s just defining the eponymous “A” and “B” in your A/B test.
In the above hypothetical example, let’s define the experiment like this: “I’ll show 50% of my visitors’ ads on the right hand side of the website and 50% of my visitors’ ads on the left hand side of the website”.
Boom. There we have it: an experiment.
Executing the Test
Now that you’ve designed your experiment, how do you set it up?
First, select “A/B Experiment” instead of “Multivariate Test”. While a multivariate test is one of the most robust ways to test your website, it’s a little more complex of a solution that we need in this preliminary test.
Remember, if you’re having any trouble, try working through some of Google’s tutorials.