In an older but still post, Steve Hanov explains a nice algorithm to test out new features on your audience. I don't now if this approach will yield better results, but it sure is an interesting idea.
In recent years, hundreds of the brightest minds of modern civilization have been hard at work not curing cancer. Instead, they have been refining techniques for getting you and me to click on banner ads. It has been working. Both Google and Microsoft are focusing on using more information about visitors to predict what to show them. Strangely, anything better than A/B testing is absent from mainstream tools, including Google Analytics, and Google Website optimizer. I hope to change that by raising awareness about better techniques.