If you liked this, you might also like that. All of us have stumbled upon this phrase on the internet. And sometimes it is true and sometimes totally wrong. While many people are very critical about the vast amount of data collected by various companies (and I must confess, it is a little creepy from time to time), in general I appreciate that my user behavior is watched.

Big data is changing the world

Questionnaires, likes, dislikes, reviews.. We have always been forced to make a statement on whether we like something or not and the market has responded to new demands. The only thing that has dramatically changed during time is the way we communicate what we want, from a direct and personal way in the past to an unknowing, impersonal way of communication. It seems as if we lost control on our own internet-personality. And hand in hand with this development also the amount of information has dramatically changed. But tech-companies took up this challenge early on and tried to find solutions to big-data related problems. The main question of companies dealing with user data was and still is:

How can we make use of these data?

Almost every business is affected from this rapid development. And it is no surprise that world-famous tech-heros like Amazon, Google, Facebook, Apple or Uber are on the leading front of the big-data revolution. They are dealing with these kind of data since their beginnnings and try to develop new tools and algorithms. One cannot ignore how these companies changed the world.

Netflix rent DVD´s?

With their worldwide-launch in 2016 Netflix, one of the internet television giants, is now also one of these global players, even if they are only changing the way we are being entertained. Think about the process classic television has undergone in the last two decades. Actually there is not that much process. That is why the idea of internet television was in some kind revolutionary. And Netflix is one of those companies that understood very early that the key to success lies in going from offline to online.
Wait. Netflix has a offline past? Yes, they have!
Netflix was actually founded 20 years ago as a DVD-renting company in the US. The major problem they encountered during their first years was, that they were not able to deliver or store the most recent blockbuster to every single user. So, they tried to change their system and thought about a way to stand out from the crowd. And they found a clever way. They started introducing a recommendation system to their customers, based on availability, genres and more features, they took into account. And surprisingly people liked it. From that point on, Netflix asked users about their preferences and collected data, therefore improving the recommendations for the next movie to be rent. The fact that Netflix nowadays is impressivly successfull with more than 100 million subscribers is closely tied to tremendous improvements of the recommendation algorithm.
But why is it even important? Let´s do a simple experiment! Think about any movie or a TV-show you want to watch in this very moment. What will it be?

Actually it is quite hard to think of anything special. At least for me. So, I really appreciate Netflix´s way to suggest content and apparently many others do: 70% of what people watch on Netflix is based on their recommendation algorithm.
That´s a lot.

How does it work and why is Netflix so good at it?

The basic idea of recommendation is:
Users that had similar interests in the past, will also have similar interests in the future.
So, in simple words, Netflix´s algorithm is based on the fact, that there are people around the globe with a similar taste in movies.
The algorithm seems quite accurate today and only occasinally nonsense-recommendations pop up. But this has not always been the case and in their first days Netflix tried very hard to improve the algorithm behind.
Every story you find about Netflix on the web mentions the famous Netflix price, a competition starting in 2006. It was meant as a machine learning and data mining competition to improve movie rating prediction. They offered $ one million to anybody improving their current system by 10 % RMSE. Already at that time they saw it as a key component of their system and to success in the future.

Why they 👎 the ⭐⭐⭐⭐⭐-system

Since then a lot has changed and now every aspect of our viewing behavior is being recorded. Location, date, time, device, pausing, fast-forwarding, browsing, scrolling, ratings. All of this user-events have an effect on what is recommended and how the surface of the app looks like.
For all of you long-term subscribers, you might have noticed, that a few months ago they changed their rating-system from 5-stars to a simple like/dislike.
Why would they do that? It is quite obvious that this comes along with a loss of information. During a process called A/B testing, where they randomly direct visitors to two versions of the website (one with ⭐ and one with 👎 /👍), they found out that users pushed the like/dislike buttons twice as often in contrast to 5-star rating. So they changed it. The reason of this big change is well axplained in numerous blog articles.

Cancer is cured by watching Netflix?

What can we learn from Netflix? As the title implies there is a value of Netflix beyond entertainment. Obviously cancer is not cured by watching TV, but companies, like Netflix and especially their data science teams had to solve big data problems in the past. With that they were and are pushing the limits of computational science. Some of their tools, like special cases of matrix factorization, have already found their way into science.
While Netflix needs those methods to find patterns in movie ratings, genomic researches try to find patterns in sequencing data. The analogy between those two totally different fields is remarkable. Computational solutions for big-data problems are one of the key components of a personalized medicine in the future, as it can be expected that each patients genetic code will be solved.

House of Cards was programmed to be a hit

But back to the entertaining part. Ever heard of House of Cards? This show changed the game and opened the way to now very popular Netflix-Original series. It was one of the first internet-TV series, the first series to release an entire season at once and the first series programmed to be a hit.
Early on, Netflix saw the potential of building their marketing model upon data and they are still following this way.
With more and more people subscribing worldwide and advanced technologies coming up, they are facing new problems. But from the past it is quite clear, that they will make the best out of these changes.
One can expect exciting things…

Interested in some more facts?

A great source for anybody interested in Netflix´s methods, algorithms and so on is their tech-blog

Is cancer cured by watching Netflix, or do I get it wrong? Elana Fertig, Assistant Professor of Oncology at the Biostatistics and Bioinformatics department of John Hopkins University, gets it right in her article

Somehow outdated, but still funny interview of two of Netflix´s recommendation crew

Read how Silicon valley billionaires influence the american school system

Or just browse to your favourite show..