Whew, so much Netflix love it takes two posts to get it all in. This time I want to talk about how Netflix leverages recommender technology to solve a classic business problem: "How do we keep them coming back?" Since Netflix is a subscription-based service, they know that when the user runs out of movies they want to watch, the subscription's in danger. And most users don't have an encyclopedic knowledge of films that can keep them renting for years. The solution, then, is to make excellent suggestions to the user so their queue never runs dry.
Netflix does this in two main ways -- with automated recommendations and with social connections to other users. The automated recommendations kick in as soon as a user loads the main page, where most of the real estate is given over to several categories with four featured movies each. The categories, which are based on films the user has rented or rated, could be genres like "Crime Thrillers" or "Spoofs and Satire," they could be regional ("Films from Hong Kong"), or they could involve specific actors or directors. Important points here are that the categories shuffle with repeated visits to the site, and that the system always explains to the user which specific films generated those suggestions. The rotation ensures that the user will continue to pay attention to the recommendations as time passes, and the explanations help the user to trust the system's picks (this is always a very big concern in recommender systems).
The social element is also a big part of Netflix's recommendation strategy, except here the goal is to have your friends (rather than the system) suggest movies for you. The community dashboard page presents a summary of friends' recent queueing and rating activity, along with user-generated thematic toplists, reviews, and "friend quizzes," which are one of my favorite parts of the site. Here, the user is challenged to guess which one of four films a friend either loved or hated - for me, this is always interesting and has lead to some cool "seriously?" moments. There's also a cool web 2.0 sliding-window interface to view friends' queues and ratings, which is a very smooth way to handle paging through large lists of movies.
Overall, the Netflix strategy is to throw movies at the user in as many different ways as possible. Is it working? Well, since the last post, my queue is up from 70 to 86. And more importantly, my queueing rate is vastly higher than my watching rate, so it looks like I'm on the hook for some time to come... good design, eh?
New home
4 years ago
