Traffic to BusSongs.com is on the up, thanks to the ever popular collection of Christmas and Halloween songs. Wishing to captialise from the increased activity, I’m stepping up efforts to improve monetisation.
I embedded an AdSense Link Unit into the navigation last May and, foolishly, haven’t really touched it since. About a month ago, I went nuts and rolled out a test to see if I can tweak the performance of the ad unit. I’ve now tried an assortment of different colour palettes, positions, titles, and behaviours. The below swatch shows all of the test candidates:
Each incoming visitor to the site is assigned to see one of the above, and they continue to see that version for the following 30-days (as to remove bias resulting from a constantly changing style). Thanks to AdSense’s Custom Channels, one can easily track the performance of each version and compare that to the original (aka the control).
I’ve now run the experiment for about the month, so the data should be pretty reliable by this point. Here are the results:
It’s pretty incredible to so clearly see how a relatively minor change can have a significant effect on performance. Test #17 was born from trying to take the best features from other high-performing ads (ie. the best of breed). I need to extend this further and run a few final tests to see if I can iterate any further.
One problem I’m facing is that it isn’t necessarily prudent to strive towards finding the best performing ad as measured by CTR. An ad that generates the most click-throughs may also be alienating users and altering their usage on the site. Therefore, I ideally want to track the behaviour of the site’s users for each of the ad versions. This doesn’t appear to be possible in Google Analytics; does anyone know how this can be best accomplished?