One of my friends, who is regrettably still in stealth mode or I would be shillin' like a villain, is working on an automated AB testing framework. Suppose you had a website with a banner, a sidebar, and body text with 3 paragraphs, each with 4 possible variations.
His software, as far as I understand it, genetically engineers the optimal web page for conversions for you. It apparently is already running on his (unrelated) software site and he credits it with helping him push through the $1 million mark. He also says that it doesn't assume there is a static optimum, but rather that different visitors might need to see different things (e.g. hypothetically, if someone has seen your screen shot page they probably don't need more visuals to make the purchase decision, so you'd use more screen real estate on textual benefits and less on screenshots).
I'm going to be beta testing it on my site, and I'll be blogging progress as usual, if anyone is interested in hearing about it. I'm pretty excited. I'm told the tracking code will be ready in about a week or two and the functional bits ready for beta relatively soon after that. (Nice rolling launch strategy, incidentally -- gives him time to collect data for the target sites and resolve the inevitable Analytics Is A Hard Problem (TM) issues, then the black magic algorithm will have some seed data for when it gets started.)
His software, as far as I understand it, genetically engineers the optimal web page for conversions for you. It apparently is already running on his (unrelated) software site and he credits it with helping him push through the $1 million mark. He also says that it doesn't assume there is a static optimum, but rather that different visitors might need to see different things (e.g. hypothetically, if someone has seen your screen shot page they probably don't need more visuals to make the purchase decision, so you'd use more screen real estate on textual benefits and less on screenshots).
I'm going to be beta testing it on my site, and I'll be blogging progress as usual, if anyone is interested in hearing about it. I'm pretty excited. I'm told the tracking code will be ready in about a week or two and the functional bits ready for beta relatively soon after that. (Nice rolling launch strategy, incidentally -- gives him time to collect data for the target sites and resolve the inevitable Analytics Is A Hard Problem (TM) issues, then the black magic algorithm will have some seed data for when it gets started.)