And what about page impressions? Were they simular in both cases? May be you'd better use some code to show half times one ad and half time few - and then results for 1-4 weeks will be much more trustable. Now I experiment at my site with 4 different adds - trying to find which one converts better - then may be I will try to turn off few ads too. Thanks for experement.
I run 2 large rectangles in each thread for guests in each of my forums. If I would remove the 2nd on my main earning forum I would lose mid 3 didgit income each month. Don't think I'll do that Page Impressions 05.-15.05.: 10401 18.-28.05.: 8371 Uniques: 05.-15.05.: 7750 18.-28.05.: 5875 The Idea for testing you have is great but unfortunatly due to the code changes required very hard to accomplish. My ads are not slapped on my site they are integrated with some extensive CSS which is hard to run/remove in a split test.
I just have one modestly sized ad on my forum. I think it gives my members a better user/visitor experience this way and shows that the site's primary purpose is not making money. People I think are more inclined to trust a site that isn't plastered with ads all over the place and therefore just as likely to click on the one small but easily visible ad, rather than see a spam and ad filled site and just click back out.
By the way it is easy to test (make a split-testing) using JavaScript - google allow this code. <script type="text/javascript"> var random_number = Math.random(); if (random_number < .5) { google_ad_client = "pub-*****" google_ad_slot = "***"; google_ad_width = 336; google_ad_height = 280; } else { google_ad_client = "pub-*****" google_ad_slot = "***"; google_ad_width = 336; google_ad_height = 280; } </script> <script type="text/javascript" src="http://pagead2.googlesyndication.com/pagead/show_ads.js"></script> PHP: You can just make it something like this. But one code parameters blank - to test. 05.-15.05.: 10401 18.-28.05.: 8371 - it's not a small differens.. So 20% less traffic made you more money - I think is good result. May be less ads is better for true.
My site is a business. However visit it and tell me if you think it is "plastered with ads" A split test is easy, I know. I do these for 5 years now with my Ad Server. But a split test with extensive html and CSS code changes to your forum, this is not easy You are right. I will test this more extensivly on my high performing site.