Netflix is gigantic, and they know it. That is not in dispute in the anonymous event venue during a drab and rainy day in Berlin when the streaming giant invited the media to explain how they work. 109 million unique users, and many times more sub-accounts – the numbers are impressive, and it is easy to see that Netflix has come a long way since the late nineties when they sent out DVD movies through mail-order. Nowadays they are not simply a streaming business for TV series and movies, but also a significant producer with a growing list of qualifications. During the last couple of years Stranger Things, House of Cards, and Orange is the New Black trickled out from the factory, to mention a few.

While the platform has grown, challenges have been added. Such as how Netflix will make all customers satisfied with the supply (and keep on paying), or how the streaming giant will succeed in reaching every single user with the right material at the right time. After spending one day here the answer is data. A whole lot of data.

Big Brother is Watching You

The fact of the matter is that Netflix knows what we do, into the last detail. They know when we are pausing a movie, and exactly how long it takes before we press the button for next episode of a TV series. The data is then used to streamline our use. Or in other words: to get us to watch Netflix more.

The interface is one example of this. If you have used the service for a while you might remember the days when static pictures in different sizes dominated. Today the material is shown in huge boxes with moving pictures. It is quite minutely consummated. After a series of so-called a/b tests (Netflix shows one interface to a group of users and an alternate one for another group), Netflix concluded that the users would stay longer if the material is presented in form of a short video clip.


This way we know that someone will love a certain series, even if they never watched it before.

Product Manager Greg Peters says that we are finding new series today just by flicking through Netflix.

“We noticed that if we created a story around The Walking Dead through a picture and centered it in the interface the user found it more pleasing. In fact, we have tried to use that kind of interface before, but then our TV sets were not good enough. No one wants to wait for the material to load. It is only lately that we have reached a point when the infrastructure is in place,” says Greg Peters, Product Manager at Netflix. He explains that each account has three sub-accounts on average.

“We take the data from each user, study what they watch, what they don’t watch, and how long they watch. This means that we know if someone will love a certain series, even if they never watched it before. We use the data to tailor that account.”

Selection through manual labor

A necessary part of the strategy is that the underlying system is getting better and better to categorize our tastes in very small compartments and then present new things to us, based on our preferences. The more specific, the better. Things that in the beginning were categorized as ”drama” is changed to “French dramas about friendship” as soon as the streaming service has gathered our habits.

Behind all this is a way of working which is quite simpler than one might believe. All around the world, there are about a hundred “taggers” – in Netflix’ own terminology – who in short get paid to watch Netflix. They go through everything minutely. Every movie and TV series are broken down into huge amounts of data points. Episode by episode, scene by scene. They do not only decide the genre for a specific movie, but also the tone of a certain scene or the town the story takes place in.

Mike Hastings, who manages this area, points out that the taggers do not imply how funny a movie is, just how many potential laughs there are. If possible there should always be at least two taggers per film and TV series. Then the risk of partiality is minimized.

We regard the parts as ingredients, sort of like cooking.

Mike Hastings explains that among the taggers there are people from a lot of different backgrounds, some are even scriptwriters.

” We do not value the content when we decode it. There are categories like acclaimed by critics, but there is, for example, no best comedy. We regard the parts as ingredients, sort of like cooking,” says Mike Hastings.

“If we have no categories at all the users get confused when they are choosing what to watch. The sorting makes it simpler. When we categorize what we offer, for example, thrillers, people get curious.”

We can look at the comedy series Master of None as an example of how it works. A simple analysis of what kind of series it is will probably land in some kind of “a slightly black comedy in an urban environment”, but in the world of Netflix, we go deeper than that. Here the taggers have assessed the tone (humorous and witty), what the story is about (dating, career), and the personality traits of the main character Dev Shah (confident, bored, funny). In addition, things like how much sex there is (some), how much drinking (moderate), the amount of violence (none) are ranked.

Master of None
This is what some of the data points for Master of None looks like.

”When we have the data, we can begin offering the users different things depending on what they have watched earlier. We want people to think: ‘Netflix understands me’. The objective is to get our users to choose swiftly what they want to watch.”

Mike Hastings also shows quotes from users who have contacted Netflix and explained that the service managed to get their taste in movies spot on.

But, it is not strange at all that, after a considerable number of hours in front of the Washington-based drama Designated Survivor, you’ll be recommended House of Cards and similar political shows taking place in the North American capital.

Netflix also localizes the categories carefully, or, as Mike Hastings says:

“Sitcom is a term that mainly works in the US, and we only use it there. It is important to get the right things working in France, Germany, and so on.”

Red boxes fix the streaming

Behind the data and flashy interfaces, the end-all and be-all is of course that the technology works. Netflix hasn’t always been in the forefront in this area. For a long time the service has worked well on our mobile phones and gaming consoles, but in our smart TVs and digital boxes, it has been worse.

Luckily that is much improved, even if it’s not optimal yet. Obviously, Netflix tries to do their share and the clearest sign is the six months old marking Netflix Recommended TV – TVs with adapted software for Netflix. This includes different criteria, among others that starting Netflix must be quick, that the latest version of the service always is used, and that the TV remembers what you did last.

The objective is that Netflix should be experienced the same way, no matter what unit you use.

Maria Ferreras explains the importance of good partners. It should be some kind of win-win situation where the manufacturer and the broadband supplier get added value from offering Netflix to the customer.

“We have worked with manufacturers from all over the world. The objective is that Netflix should be experienced in the same way, no matter what unit you use,” explains Maria Ferreras, Business Development Manager at Netflix.

To get everything flowing as smoothly as possible Netflix has, since a couple of years back, installed red boxes in the Internet suppliers’ stations. Here the material is stored locally to lower the technical disturbances as much as possible. The higher quality we demand, the more important this becomes.

Red boxes Netflix
Boxes like these are placed in a number of stations. The closer to the user they are, the better.

”If we hadn’t had them we would have needed to stream the material from our main servers. It would have been difficult to get sufficient quality in Europe if the material is sent from California.”


Are there any risks with all the data, that Netflix’ logotype soon will sit on every TV, and that the company’s resources are so great that they are becoming difficult to compete with? Except for the obvious integrity issues, there is, of course, the supply. When everything can be cut down to data points the risk is that the unique will disappear in favor off lavish but streamlined productions, directed at as many as possible. In that context, Netflix’ role as producer isn’t completely unproblematic. Huge amounts of money are spent on Netflix Originals – the label for their own material and some exclusively licensed material – and it needs to yield a good return.

Product Manager Greg Peters:

“The interest in Netflix Originals is huge. A lot of people want to watch this material.”

At the same time, it gets a lot of exposure. It is hard to miss that you have a new season of Stranger Things out right now.
“Yes, it’s true. But we always try to find a balance between Netflix Originals and licensed material. Obviously, our own material is important to us, but we don’t want to advertise it too much, or offer things the users aren’t interested in.”


It’s all about the data. Everything needs to be measured, collected, and analyzed. And what the masses don’t want will disappear, preferably without us noticing. When we leave Netflix that day we feel that they know almost everything about our TV habits and aren’t afraid to use that knowledge. If Netflix wants us to watch something, we can probably trust that it will be shown.