A Quirky Netflix Recommendation
I've been very impressed with the quality of recommendations Netflix has offered me. So much so, that it would take me three years to watch all of the movies in my queue if I maxed out my monthly DVD deliveries and online streaming.
The recommendations are so good that bad ones really jump off the page. For example, I loved the movie Lost in Translation. Based on user behavior, Netflix recommends the following films:
- Sideways
- Rushmore
- Garden State
- Syriana
- Being John Malkovich
That makes sense. But what about this?
I'd like a snowboard movie called "Lost in Transition" because I liked "Lost in Translation?"
Swing and a miss.
But the fact that this was noteworthy to me shows that Netflix is doing a lot of things right.
1. Posted by: Jeff Deitch on September 23, 2007 10:43 PM:
It may be the case that people who liked "Lost in Translation" may have been confused when they saw this and said, "yeah, I really liked that movie. Bill Murray is great!" and given it a good rating. Being that you and these people both like "Lost in Translation" you probably have similar overall movie tastes. Netflix sees this and because these others said they really liked "Lost in Transistion," Netflix recommends it for you.
I haven't really thought about that, but that is an interesting problem. What happens when Netflix users accidentally rate movies they haven't seen? Does Netflix look for outlier data from its users and throw it away assuming it was a mistake? Or do they just live with it and accept that their recommendations are then flawed?