If you use Netflix regularly to stream movies and shows, then there's a really good chance that its recommendations play a pretty big part in your viewing habits. This will likely be even more the case from now on, now that the user profiles are rolling out. Now, you won't have the distractions of what Netflix thinks other people in your house want to watch. It's going to be more personal than ever.
Wired spoke with a couple of Netflix engineers, producing a rather interesting look into the kinds of things Netflix takes into consideration when determining what to show users as recommendations. It turns out, as you might have guessed, that they use a lot of different data points related to your usage habits. I say usage, because it's not just about viewing. In some cases, it's literally about how you interact with the Netflix interface (in addition, of course, to your viewing habits).
"We know what you played, searched for, or rated, as well as the time, date, and device," explains engineering director Xavier Amatriain. "We even track user interactions such as browsing or scrolling behavior. All that data is fed into several algorithms, each optimized for a different purpose. In a broad sense, most of our algorithms are based on the assumption that similar viewing patterns represent similar user tastes. We can use the behavior of similar users to infer your preferences."
He also says Netflix is working on incorporating viewing time data into the recommendation algorithms.
"We have been working for some time on introducing context into recommendations," Amatriain tells Wired. "We have data that suggests there is different viewing behavior depending on the day of the week, the time of day, the device, and sometimes even the location. But implementing contextual recommendations has practical challenges that we are currently working on. We hope to be using it in the near future."
Another interesting bit of the interview has Carlos Gomez-Uribe, VP of product innovation and personalization algorithms at Netflix saying that "predicted ratings aren’t actually super-useful." This, of course, was what the famed "Netflix Prize" was based on.
Funny how things change.