Img source:

How Netflix Uses Big Data

Netflix doesn’t just use big data, it harnesses it with spectacular efficiency and accuracy.

Just consider this: in 2006, Netflix announced a competition for coming up with an algorithm for predicting what movies a user would enjoy, based on his/her movie preferences. Not only did they pay the winner $1 million, but the algorithm was a success – it improved their recommendation system by 10%.

However, to everyone’s dismay, Netflix never actually used it, as “the additional accuracy gains that we measured did not seem to justify the engineering effort needed to bring them into a production environment.”

This just goes to show how high Netflix’s standards are when it comes to using big data efficiently. Another even more eloquent example of that, which we’ll explore in more detail below, is the $100 million-investment, unprecedented for Netflix, which they in the series A House of Cards, based almost entirely on data rather than plot.

How Netflix Obtains Big Data


According to CNN, Netflix now has 139 million subscribers, and it makes each and every one of them count.

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

What he means by that is one user’s experience can be entirely different from another’s, and that’s because Netflix’s goal is to tailor that experience as closely as possible to the individual viewer’s preferences. The closer the match, the happier the user, the longer he’ll stay with the service.

How Big Data Works for Netflix Compared to Television Networks

Professor of Big Data at the JCU Online explains, “To give you a slightly abstract idea, let’s say both Netflix and TV networks have a “persona” in place that represents their average customer, a persona they strive to cater to.

While TV networks might have an idea of that persona’s preference in genres and when is the persona more likely to watch, Netflix knows the persona’s favorite actors, directors, settings in terms of time periods and locations; preference in programs’ age; when the persona watches series and when films; how much time does the persona spend watching per sitting, when does the persona rewind or fast-forward, etc. and etc.”

In short, because Netflix is an online service, all its data is exact rather than based on samples of viewers, unless we’re talking about surveys, and even then, the results are far more accurate since the platform has much more reference points and noticeable trends.

Once the Credits Start Rolling…


The analysis of big data has let Netflix in on a little, big secret: their viewers are likely to disengage from the platform once the credits of the film/series/program they’ve just watched start rolling. They might feel like having a break, doing something else, and a hundred of other things outside Netflix.

So, once the credits start rolling, the battle for viewers’ fleeting attention begins. If Netflix manages to re-capture it before it flies away to something else, their subscribers will ultimately spend more time on Netflix. And the more time they spend on it, the less likely they’ll be to unsubscribe.

In fact, Netflix most likely has quite a specific target for the amount of hours their subscribers need to spend on their platform in order to be the least likely to unsubscribe.
So, how does Netflix get viewers to keep watching and even engage in the infamous binge-watching?

Auto-Playing Next Episodes and Recommendations


Those two features, like any Netflix feature for that matter, are anything but random.
When we’re watching series, there’s always this delicate moment of hesitation, when everything hangs in the balance. You’re asking yourself the ultimate question: “To watch, or not to watch another episode?”

Netflix’s auto-play feature capitalizes on this moment and does its best to take the choice away from you. Before you know it, the next episode is already playing. It’s like it’s almost meant to be. It may sound too simple, but in many cases, that’s all it takes to get us to indulge.

Research has even shown that we’re inherently wired to binge-watch, all we need is a little nudge.

And even if viewers openly intend to watch another film or ha, they would normally have to start searching for another movie/program. The process is often long, arduous, and often even worse – thankless. After a while, users might give up and just do something else. If that happens a few times over a certain period of time, they’ll be much more likely to cancel their subscription.

Here is where Netflix’s recommendation algorithm comes in, and considering that roughly 75% of Netflix viewing is driven by it, it definitely seems to be doing a good job.

One of the small, big ideas, credited for the algorithm’s success rate, is the implementation of the thumbs up/thumbs down rating system in place of the 1-5 stars rating system, in addition to the myriad of other factors that weigh in the recommendation, like the ones, mentioned above, and the ones that you can read below.

The House of Cards Case


As mentioned in the beginning, Netflix made an unprecedented for them investment, outbidding the likes of HBO and AMC, for the rights for a U.S. version of House of Cards. Spending over $100 million for 2, 13-episode seasons, this was a huge business move for Netflix, but that’s precisely what it was – a highly strategic move, not a risk.

Investing in an original series is always a risk – there are so many factors that interplay to determine success, like the script, the director’s vision and execution, actors’ performance, etc. But those factors didn’t concern Netflix one bit.


Instead, they looked at the data. First, the British version of House of Cards was a success. More importantly, a lot of Netflix’s subscribers:

– had watched The Social Network, directed by David Fincher who also directed House of Cards, from beginning to end.
– who had watched the British House of Cards also watched movies with Kevin Spacey and ones, directed by David Fincher.

Steve Swasey, VP of Corporate Communications, says it all in an interview for Gigaom:

“We can look at consumer data and see what the appeal is for the director, for the stars, and for similar dramas.”

House of Cards went on to becoming a huge success and has come to epitomize the tremendous power of big data when used right.

Peter is a freelance writer with more than eight years of experience covering topics in politics. He was one of the guys that were here when the started.