If the term split testing is new to you, the basic idea is that you run ads two different ways and see which one works best. “Works best” can mean whatever you want it to mean…best click through rate, best earnings, etc.
My plan had been to do serial split testing. Run ads the way I have them for a two week period, record the click through rate, impressions, and earnings, and then run them with different colors for a two week period. I would wait to change formats until a later test, since one of the critical parts of split testing is that you must know what causes any change in results. So if I changed both the color of the ad and the format, and the click through rate went up, I wouldn’t know which change caused it. It might be that it’s the different format, and the changed color is actually hurting my click through rate.
Change one thing, test it, and see what the results are before changing anything else.
Another option is parallel split testing. This is where you run both sets of ads at the same time, rotating equally between them for all your web visitors. Visitor #1 might get blended ads, visitor #2 would then get ads of a different color, visitor #3 would get blended ads, and so on. I prefer parallel split testing because it equalizes any possible seasonal factors. For example, my two week period for blended ads started on March 11th, and will end March 24th. March 24th is the day after Easter here in the United States, and that might have an effect on web traffic and click through rates.
By doing parallel split testing, both ad variations would be running at the same time.
One of the tricky parts about parallel split testing is that you really want each visitor to see only one ad variation. A simple rotation scheme on a page load won’t do that, since any given visitor might reload any given page on your site. If they reload it they might see the other ad variation. Changing ad colors on a visitor will get their attention, and will definitely skew the results you’re trying to measure.
Each visitor should see a consistent ad theme on your site, and it should rotate on a per-visitor basis.
I’m still hunting for a suitable script to do this for advertising on a WordPress blog. There’s a split testing plugin out there that is for Adwords campaigns leading to your blog, not for Adsense on your blog. There’s another one for split testing different variations on your posts. That one looks very interesting, but wasn’t what I wanted.
I did find the most excellent Who Sees Ads? plugin for managing ad display. The plugin will allow this with some custom PHP programming added. With the plugin, you can define rules under which visitors will see any particular ad. You could display ads only to new visitors and not to your regular readers, you could display one ad to visitors from the Unites States and another to visitors from India, you could specify that each visitor should see an ad only X times, etc.
One of the options in the plugin is to use a custom PHP function to determine if an ad should be shown or not. I think, with some custom PHP code, that I can manage to get a PHP function that will alternate between displaying ad variations with the plugin on a visitor basis, and not a page load basis.
The only problem is time! For the meantime, I’ll stick with serial split testing, and begin my two week period of testing non-blended ads tomorrow.
Look back in two weeks for the results.
Update: I just found the WP Spinner plugin. It’s basically a plugin that will rotate different content on your WordPress site. They say it can be used for split testing, although I’m not sure if it’ll show the same content for a particular visitor, or if it’s a simple rotating on page views mechanism. I doubt I’ll give it a try, since the programmer in me wants to write some PHP code for use with Who Sees Ads, but you might check it out if you’re looking for split testing plugins.