- 10/06/2015


Recommended for You: How Netflix Uses Data to Keep You on Your Couch

Netflix has over 50 million members, myself included. And they know a lot about us. No, they don’t know that I had a delicious chicken parm for lunch today, or that I’m technically on a Paleo diet but I can’t muster the strength to deny myself a hot, saucy sandwich from time to time. But they do know that I’ve been watching old episodes of 30 Rock after work. And that I watched the entire first season of Orange is the New Black over the course of two days (I think it was raining that weekend?).

What I’m trying to say is, Netflix has a lot of data. Not only are they tracking what we’re watching, but also when and where we’re watching, whether we’re watching full episodes or pausing and coming back to them. They track all these things because they know how valuable data is. How, if used well, data can improve their product, their customer experience, and of course, their bottom line.

The best example of this is the multitude of algorithms Netflix uses to create personalized homepages for each of their users. Back in 2006, Netflix held a machine learning/data mining competition to improve their movie rating predictions. Rating predictions then evolved into a personalized ranking of Netflix’s entire catalog.

Now, they’re developing personalized page generation – not only is data driving the recommendations Netflix gives us, but also the presentation of those recommendations. The movies grouped into each row, the order of those rows, even the names of the rows are data-driven. And it’s working – 75 to 80% of the videos Netflix users watch come from these personalized recommendations.

Netflix has said that their data-driven culture is key to their success. But of course, it’s not just about the numbers. At the end of the day, someone still has to be there to make the big decisions, such as which movies to license in the first place. There’s no doubt that Netflix weighs viewer data against licensing costs, but that’s not always enough to predict a hit. Netflix’s chief content officer Ted Sarandos uses a “70/30” approach, (70% data, 30% judgement), for commissioning decisions.

To get the best mix, Netflix knows you have to use your data well. Simply having lots of data points is not valuable in and of itself. Rather, it’s how you approach that data, how you analyze and apply it, that will be valuable. Netflix’s data team actually outlines their data philosophy, which sums up nicely how they approach their large and presumably complex datasets.

  • Data should be accessible, easy to discover, and easy to process for everyone.
  • Whether your dataset is large or small, being able to visualize it makes it easier to explain.
  • The longer you take to find the data, the less valuable it becomes.

Netflix had over $5.50 billion in revenue last year. You’ve heard it from the best. Now go on and see how you can put your data to good use (do some research on Netflix, we won’t judge).