Why Netflix Knows What You Want to Watch—It’s Math

Author: Priya Chhatwal (M.Sc. Statistics and Data Science)

Do you know what that feels like?

You watch a late-night show and then boom! Netflix is already giving you the next perfect binge: a show you didn't even know existed but now can't live without. It's not a conspiracy, and it's not magic. In fact, it's not even that hard. It's just some smart math working in the background.

We will call that math wizard SVD, which stands for Singular Value Decomposition. The idea may sound hard, but it's a lot more interesting than it sounds.

Consider yourself at a food festival where there are a plethora of different vendors. Even though you haven't tried much, your friend, who is an absolute foodie genius, is aware of your preferences. You say, “Hated that ramen, loved that pizza." And based on that small piece of information, they are spot on: "You'll love the lasagna, but you should probably skip the pad thai."

That's SVD in a nutshell. Using the little information it has about your preferences, it determines what else you're likely to enjoy by examining the work of thousands of other users.

The recommendation system operates behind the scenes as a huge table. Songs or movies on the one hand, and users on the other. You haven't seen everything yet, so the majority of it is blank. The whole point of the game is to fill in those blanks.

The key component is SVD. It discovers your taste's "hidden layers." Consider it this way: Every film has a secret formula: a small amount of action, a little romance, a little sci-fi, and a little emotional depth. You have a distinct flavour profile as a user as well; for example, you may be 40% action, 30% romance, and 30% humour. It is SVD's responsibility to figure out both recipes. It's like, "Hey, this new movie has a recipe that matches your taste profile perfectly!" once it has them. Before you even have time to consider what to watch, it recommends something.

The most well-known instance dates back to 2006, when Netflix held a significant competition to enhance their recommendation system. "Oh, people who liked Inception also liked Interstellar," was not the only statement made by the winners. It's too easy. Rather, they employed SVD to delve deeper and discover the reasons behind the popularity of those films—the intricate storylines, the surreal themes, and the high stakes. In this way, even if you had never heard of Arrival or Ex Machina, they could recommend them. It's the art of discovering the more profound connections.

Furthermore, this isn't limited to Netflix. Similar techniques are used by Spotify for customised playlists. Amazon is aware of the devices that consumers may desire. It's even used by dating apps to link users. "Trust me, you'll like this," says SVD, who connects people.

Remember the math the next time Netflix recommends the ideal weekend binge. It's neither a crystal ball nor luck. It's SVD performing its mathematical magic in the background. Having a little math on your side is helpful in a world full of options.