I agree that one day might be too little to test but on the other hand I will argue that if you test something that is radically different and you see a steep decrease over the course of that day you shouldn’t wait for several days or a week to change it back. If something clearly doesn’t work it should be instantly removed.

You can’t just say you are going to sample over X visitors or Y impressions, because they vary greatly with the time of the day, and day of the week.

You must also remember that changing the layout of the ads will initially give you a spike, so you should probably wait 1-2 days before recording results.

Normally though it is not a matter of “time” but rather a matter of the number of visitors taking action.

e.g. you change every time you have had 100 clicks or so. That is what I would do.

On my site I have about a 3/4% CTR, which kinda sucks considering I have the exact placement mentioned on adsensemasterplan. I think people generally know what they are looking for when they come to the site.

I am, however, currently toying with the idea of a self learning adsense placement system. That basically tries new placements and records the CTR and eCPM. It’d probably take a few months to work, but would eventually give you the best return for your design.

Jumping from one thing to the next rarely helps in the short run.

And thank you for commenting. Please come back again soon.

Mikael