by Carlos Portocarrero, WeSeed Writer
Netflix (NFLX) has won over millions of people by thinking outside the mailbox. DVDs by mail? Brilliant! Besides convenience, Netflix also has a huge library of DVDs. But that’s good and bad: Choosing between that many DVDs can be pretty overwhelming for a Netflix user, which is why their movie-recommendation feature is so useful and popular.
And it’s going to get even more popular—Netflix is about to pay out a $1 million prize to a team that supposedly improved the movie-recommendation algorithm by at least 10%. The idea is to keep you away from the duds and serve up only the movies you’re really going to like.
Using their own proprietary software, Netflix currently analyzes the movies you like and shows you movies you might like. Right now, Netflix is recommending Memento for me. Why? Because I liked Fargo, Lost in Translation, and Trainspotting.
I’m not sure what those movies have in common, but that recommendation is spot-on: Memento happens to be one of my favorite movies of all time. So there.
Enter the Netflix Prize, which was started in October 2006. Almost three years later, a team called BellKor’s Pragmatic Chaos has announced they cracked the case—improving the algorithm by 10.05%. That’s what happens when you put engineers, scientists, and machine-learning experts in a room together—they come up with new ways to code stuff like this.
That’s great—but is this really worth $1 million of Netflix’s money?
According to the New York Times article, Netflix has said “$1 million would be a bargain price for an improved recommendation engine, which would increase customer satisfaction and generate more movie rental business.”
If I were a shareholder, I’d be a little wary that a 10% boost of an algorithm is going to do anything for the bottom line.
This whole concept of telling people what they might like is pretty popular these days. Amazon (AMZN) has been doing it for a while with their “Customers who bought this product also liked…” feature. Pandora has taken over the music world by analyzing songs and making recommendations. Even Apple (AAPL) got on the bandwagon recently with their Genius button, which essentially does what Pandora does on a larger scale.
So sure, companies are using it. But the big question becomes: is it worth $1 million?
Image by Marit & Toomas Hinnosaar
![Reblog this post [with Zemanta]](http://img.zemanta.com/reblog_e.png?x-id=d0258570-0913-4b4b-9c77-dc23ed734451)
Whether an “improved” algorithm will actually improve Netflix’s end product is a very interesting question. Does the main question from the contest even map to the most interesting questions concerning recommendations? MediaUnbound is tackling these questions in a series of articles over the next 27 days called Countdown to 10%. You can find the series here.
Hey, do remember that even if it does not bring profits just from the recommendations, which I doubt it will, they will have developed software that makes usage easier, and at the same time, since this is a popular feature, they can sell it to make some cash off of it. Selling movies frontend and selling code backend.
Ryan, you make a solid point. This is probably one of those things that Wall Street analysts might look at and say: “this won’t have a good ROI*—bad move for this stock.”
When in reality it’s about giving customers something useful that makes the Netflix experience more enjoyable. And that focus on the customer is something that, in the long run, will benefit a company.
*ROI: Return on Investment