Split testing allows you to try different versions of your website to find the version that gets you the highest click-through rates and highest earnings. How do you set up a split test with Google AdSense? Here's how.
Determine What to Test
The first step is to determine what you want to split test.
You can either test one variable at a time, or you can test them all at once. The less traffic you have coming to your website, the fewer variables you'll want to test at once.
For example, you'll probably want to test ad positions and ad sizes before you test ad fonts and ad colors.
Once you've picked what you're testing, it's time to set up your ads.
Setting Up Your Ads
Create two or more different versions of your ad. Each different version of the ad needs to have a different ad unit, even if the code doesn't need to change.
For example, even if you're just changing where you're placing the ad (which doesn't change your AdSense code, just where you put it), create a separate ad for it anyway.
In each ad, tag it with as many relevant channels as you can. For example, tag it with the location you're putting it in, the image size, the color, etc. Track all the variables that you want to test now and even variables that you might want to test at some point in the future.
Install a Rotator
If you want to get a more complex solution, try using Google Website Optimizer. Google Website Optimizer will rotate through completely different versions of websites for you. It's a little more involved to set up than copying and pasting some PHP code, but offers more customization.
Tracking and Interpreting Your Results
Wait until you have a fair amount of data, then plug your data into a statistical significance calculator. You can find many free ones on Google. These calculator will essentially tell you based on your data how likely one placement is to win over the others in the long run, based on your short-run data.
Wait until you have at least a 90% statistical chance of being right, then call the winner and move on to the next test.