Netflix has been running their own X-prize ('Netflix prize'), a crowdsourcing competition for a million dollars to come up with a better algorithm for predicting movie preferences than the algorithm Netflix is currently using. That approach, Cinematch, currently captures 2 million new ratings (between 1 and 5 stars) per day, and is used to predict preferences ('if you liked Braveheart, you will also like...') for the 65,000 movies they rent, about 1 billion times per day. They are looking for a 10% improvement in accuracy, which doesn't sound like much, but would mean 10M of ther customers would get better matches each day - which is quite a lot, if you're trying to differentiate yourself to apture the off-line (and soon, online) rental market!.

Today, their algorithm is accurate within about one rating point of predicting how people will actually rate a movie, after they watch it. Netflix clearly sees this as both a competitive advantage, and something their customers want. In fact, it's worth a million dollars to them, and they figure it's cheaper for others to solve the problem for them than it is to solve the problem themselves.

Lots of mathematics 'experts' are in the game -  BellKor, an AT&T research group, Dinosaur Planet, a team of Princeton alums, lots of academics.. Along comes Gavin Potter, a 48 year old British psycholigist, who is using behavioral economics. He hasn't won, but has been competitive with the best teams of mathematicians from universities and industry. He even has his own consultant - his daughter, a high school senior. I'm always rooting for the underdog!

Innovations can meet or beat traditional approaches, when they are able to make connections that others haven't been able to make. The connections tend to require either different inputs, or different ways of looking at data. Sometimes, it just pays to re-think the problem, rather than using brute force.