Tag Archives: recommendation engine

Netflix understands the strong ROI of improved customer satisfaction

After three years, the Netflix Prize competition is coming to a triumphant close. This is where the online DVD rental company offered a $1 million reward to anyone who could improve its flawed Cinematch recommendation engine by at least 10 percent. Back when it started, I suggested one novel way that a competing team might improve results (hire a philosopher). We may never know all of the tricks employed by the likely winners.

And who might these winners be? A little over a week ago, the team called “BellKor’s Pragmatic Chaos” delivered a 10.05% improvement. The Netflix Prize competition has now declared “last call.” The other teams have thirty days to improve on the winning algorithm.

Two things strike me about this competition. The first is how difficult it is to predict our tastes in films. I’m frankly amazed that anyone is taking the prize. (Remember, teams have been trying for three solid years!)

The second and more important take-away is this: You can never be content with your present efforts to satisfy customers. They can always be improved — and they should be improved. Even when the cost is surprisingly steep.