Netflix’s Algorithm Toes the Line Between Artistry and Technology

A look into the science behind the endless entertainment we consume from the popular streaming service.

Poster for "Tidying Up With Marie Kondo" one of Netflix's recent hit shows. (via Facebook)

Sometimes it feels like I’ve watched every good series available, and then Netflix comes out with a new original. So, how does the streaming platform churn out addictive shows so regularly? You, me and the nearly 150 million subscribers are to thank. Big data, a term you’ve probably heard thrown around a bit, is the other not-so-secret weapon to their success.

Movies and TV shows on Netflix are tagged for various attributes — visuals, violence and valor are just a few. The age of genres is behind us; it’s all about microgenres now. Viewer demographics are not even considered in recommendations.

“You might think if we know someone’s a 17-year-old guy vs. a 67-year-old woman… you know exactly what they want to watch. Show the 67-year-old woman lots of things with Meryl Streep and Shirley MacLaine. Show the 17-year-old guy lots of Marvel’s superheroes, lots of explosions. You’d be wrong,” Netflix’s VP of Product Todd Yellin said.

The streaming service wants to know why you specifically watch a particular show or film. Is it for escapism, exploration or perhaps social currency? Did you watch “Bird Box” because you thought it would be good or because everyone was talking about it? Netflix’s algorithm records what you watch, how much you watch and what devices you watch on. This implicit data proves far more successful in predicting the types of shows viewers will want to see rather than explicitly asking.


With so many subscribers comes a mountain of data which the platform utilizes for the users’ benefit. Rows, Netflix’s main method of organization, offers tens of thousands of categories ranging from “Recently Added” to “Corporate Corruption.” Even the “Trending Now” category is customized. These suggestions are effective because they are specific to each of us. While my top picks include “Tidying Up with Marie Kondo” and “Trevor Noah: Son of Patricia,” my boyfriend’s list contains “Friends” and “Sex Education.”

“There are 33 million different versions of Netflix,” Netflix’s Director of Global Corporate Communications Joris Evers said.

Specific user recommendations is not a feature limited to Netflix. We see it in Discover Weekly albums on Spotify and selected advertisements on Facebook. But the platforms share the same problem. This dynamic programming limits the titles we are exposed to, reinforcing the options we know we like rather than offering options we don’t yet know we will like.

When Netflix uses big data to predict the success of an original series, what happens to the inherent uncertainty of creation? Netflix deploys its algorithm when deciding on potential pilots. Take “House of Cards,” for example. Netflix bid $100 million and committed two seasons before the first episode was even made. The series was engineered to be a hit; attracting fans of the U.K. original, and also of Kevin Spacey and of David Fincher. Think of big data as steroids for broadcasting: secretive, lucrative and addictive. Some of the best shows have arguably been products of this personalization technique, but what will this mean in the long run, once every streaming service has adopted the same methodology?

On one hand, it has translated to more niche entertainment options like “Black Mirror” or “Maniac,” shows that are expected to attract certain audiences, rather than generalized, simplistic series that satisfy the masses — though Netflix has produced the latter as well. It is hard to remember Netflix before online streaming and seemingly impossible to imagine a time limited to seven channels. The cable industry was disruptive, introducing dozens, then hundreds of channels. Not to disagree with “The Matrix,” but I’ve never found choice to be a problem. The more choices, the better!

Whether or not you agree with the mechanical aspect of using an algorithm to create entertainment, it’s undeniably effective. There’s a graveyard of shows canceled too soon. Big data and billions of dollars allocated to content curation prevents Netflix from making the same mistakes as so many networks. “Arrested Development” fans have the algorithm to thank for the shows return. It seems like everyone and their roommate is waiting for Netflix to reprise “The Office;” and if the algorithm leads to that, I can’t complain.

Email Skylar Carroll at [email protected]