Straight from the horse's mouth this ... A/B testing is when you show one ad format half of the time, and another ad format the other half of the time, to help decide which performs better via using custom channels. But you cannot easily do this without tinkering with the code. But now it seems like it has been Ok'd by the Adsense Team. In general you'd implement it in PHP, using a random number generator. The code looks a bit like this: More at: http://adsense.blogspot.com/2006/08/abcs-of-ab-testing.html
Yeah that's right, previously the other way to do this was to manually edit your page's code. And it wasn't randomised - this way, it may even help combat ad-blindness. Wonder what's Google's point of view with regards to using this as a permanent ad rotator, so to speak.
It could be possible by date as well.........I mean that one ad is on even dates and onther is on odd dates.........and lots of logic also available depend on logic used
rahulm: oh yeah. the possibilities are endless. I wonder if Google has now opened the pandora box with this. Though, at least this kind of code modification would not induce false clicks or change the way the ads work. So maybe they might be more lenient with this? Hmmm.
The resulting code, the code sent to the visitors' browser, is exactly the same as the code you'd get from the adsense tool. Whether you do this -by hand- on an hourly basis or some other crazy schedule, or automatically is probably trivial to google. My guess is that they don't like you tinkering with the endresult, but serverside you can do just about anything you want with it; which includes inputting php variables for colors sizes and whatever to do your testing. (<-- I have read this somewhere else too, but can't remember where...)
gandalph: ah that makes sense. But when even the ppl on the Adsense blog had to check with their "policy folks", you can see why us plebs are a bit unsure.
Pretty neat idea. I still don't think you can make a completely valid comparison between ads to judge which is better performing on a given page.
I think you need to brush up on your statistical skills... When out of 100 visitors one ad is clicked once and the other 99 times, which would you say is the better performing? And suppose your previous ad was the one that got only one click? Congratulations; by implementing the new ad you have improved your CTR by 10.000%... That's the power of knowing your metrics
Has anyone got any results from doing this then? If you have, please post them... I've set up a test on 300 pages of mine, all using A/B testing to see if black backgrounds with white writing are better than white backgrounds with black writing. I'll let you know in a week.