One answer is A/B testing, or the practice of publishing two variations of a Web page and measuring which one performs better.
A/B testing has been used by large enterprises for years. Amazon uses it to optimize product listing pages. Google was once criticized by an outgoing designer for relying too heavily on data-driven methods like this.
In general, A/B testing involves segmenting site traffic into different groups and showing each group a different version of the same landing page.
Perhaps one page will contain different headline text or a different call-to-action button style. Another page will serve as the "goal" for the desired conversion.
For example, a newsletter sign-up confirmation page or the final check-out page in an e-commerce shopping cart. The software will figure out which version of the page resulted in more clicks or conversions.