Ever since I started working professionally with web development, I've been interested in how we can improve the user experience by learning from the current users / visitors behavior. How it's possible to turn raw statistics into actionable insights - and maybe even optimize their experience automatically.
One of the first companies I worked for was Mondosoft - a site search vendor. There, I was part of a team that spearheaded our efforts into search behavior - already back in 2001 - and launched a product called 'Behavior Tracking' that (among other things) automatically would learn synonyms and common spelling mistakes from the users - and improve the search from that.
Later, within Episerver, I've spend a lot of time experimenting with (and making prototypes of) various behavioral mechanisms that aims at improving and personalizing the online experience based on behavior. For instance with content recommendations, automatic visitor segmentation and the Self Optimizing block that does real-time multivariant testing and automated improvements.