657-admin - 12.06.18

For your eyes only: How Netflix knows your taste (and sometimes doesn’t)

It used to be so simple. We watched movies and television series based on recommendations from friends and the occasional critic. The networks were major influencers, too. If it was shown in prime time, it was what the family gathered to watch.

Our viewing habits have changed dramatically. We watch whatever we want, when we want. So who influences what we want to watch? The old linear tv channels still play a part. But there’s a new player in town. Artificial intelligence now analyzes what we watch and makes recommendations.

90 Second Window

Netflix is the largest streaming service for films and series, with 125 million subscriptions – according to the company making for a total of 375 million viewers due to multiple users per subscription. The service’s content offering is massive. Last year, Netflix spent $6.3 billion on original and acquired content. Still, if viewers haven’t found anything interesting to watch after 90 seconds of browsing, they tend to switch off.

This means that Netflix spends a lot of time and money to identify the best films and series – for your eyes only. Their goal, after all, is to make you spend as much of your time as possible streaming their content. That is why the recommendation function is so vital to the company – and its programmers as important as its star actors. More than 80 percent of what subscribers watch has been recommended, Wired writes.

Your Very Own Taste Community

So what do they do to ensure that you find something interesting to watch in a matter of 90 seconds? The answer, in three headlines, is big data, machine learning and so-called “taste communities”.

The company gathers massive amounts of data about its users’ viewing behavior, then uses machine learning to analyze the patterns. In addition, a small group of employees and freelancers watches and categorizes all content.

When this is done, Netflix has a few thousand taste communities that share common tastes, and into which every user is placed. What group you’re placed in affects recommendations, the genres turning up on your front page, and the sequence of recommendations. If you are prone to romantic comedies, that genre will turn up the most often on your bespoke front page, although not being all you get.

Welcome To The Algorithm

The algorithm tests you. It wants to show you that Netflix has more to offer than Meg Ryan and Hugh Grant. It offers you a few drama series to see if you bite. Sometimes, the algorithm has a hit. Other times it misses the mark. In which case it tries to show you something else. Horror movies, perhaps.

Chris Jaffe of Netflix has told Business Insider that the company doesn’t have finacial incentives to get subscribers to watch their own original content. Unlike for instance Spotify, the production companies are not payed per view, but with a flat fee.

Netflix’ personalized front page has changed a lot the past few years. Hundreds of tests are done every year involving 300.000 customers to improve the page. They often run so-called A/B tests, where developers test two alternatives against each other.

In this work frontend development is quite important. Netflix tests everything from fonts and image sizes to description lengths. Short and sweet works best for now. Everything to keep you stuck to your screen for as long as possible.

A version of this article in the Norwegian was published on Noroff.no. Noroff is a Creative Industry Brief partner. Noroff students produce signed content for this publication through the student content agency Untold Editorial.