When I saw news coverage of Netflix’s personalization contest (they’re offering $1 million to anyone who can improve their movie recommendation system, and giving access to movie rating data to make it happen), my first thought was, “That sounds like an IMDb mashup waiting to happen.” But, Greg Linden beat me to that thought (curses!). So I’ll pile on and say, I’ll bet the winner isn’t the one with a better algorithm, but instead the one who figures out how to join Netflix’s data with some other good set of web-accessible data.
Netflix is a weird combination of great ideas and new technologies (for instance, their “instant rating” five-star system seemed like the first mass-market example of what we now call Ajax) and strangely anachronistic features (their recommendation system was bad for years, and for a long time queue reordering was like solving a Rubik’s Cube). In this case, I’m surprised that the ratings database is a download rather than an API. I’d bet that if they made it an API, the $1 million dollar prize wouldn’t be necessary at all. Going back to Greg Linden, that might get them better results.