How Netflix Knows What You’re Gonna Watch Before You Do
75% of all content that Netflix users watch is a suggested title. But how does Netflix have such a high success rate? Netflix’s recommendation algorithm is highly valued, they spend 150 million dollars per year and have 30 engineers working on it. The reason why it is so important is because their primary goal is to have you watching for as long as possible. From Netflix’s perspective, their competition is all other sources of entertainment. As told to Business Insider the Netflix VP of Product Innovation said “The biggest challenge for Netflix is: if you’re tired and it’s the end of the day, you could read a book or a magazine, you could go on Facebook, watch linear TV, or watch Netflix. We want to make Netflix so engaging you keep choosing it.” This means that they cannot just provide content for one situation. They need to have diverse content to match the consumers variety of moods.
The recommendation system all starts with data. Netflix keeps tabs on everything the user does so that their recommendations can be as personalized as possible. Their recommendations evolve with time, the more that you are using Netflix the more data their algorithm has to understand what you want to watch. The data can be split into two categories. Implicit and explicit data. Explicit data is what you tell Netflix directly. An example of this would be their ratings system, which has changed over time. They previously had a 5-star system where users could also comment their opinions on titles. Now Netflix has switched to simpler thumbs up and thumbs down system, similar to Youtube. The other part is implicit data, which is behavioral. Xavier Amatriain, engineering director at Netflix, describes how they record implicit data by saying, “We know what you played, searched for, or rated, as well as the time, date, and device. We even track user interactions such as browsing or scrolling behavior.”
This data can be further personalized with Netflix profiles. Netflix found that since accounts are used by households that usually have a wide range of ages, the content was not as well personalized. To combat this Netflix introduced profiles. Right at login Netflix users can pick their own personal profile.
These are all important because what a Netflix user wants to watch will change based on these factors.What you want to watch on a Tuesday night after a long day of work will be different than on Christmas day. What a teenage girl will want to watch will be different to what their Dad wants to. Netflix is sure to change their recommendations based on these variables. All this data is then put into a recommendation algorithm.
The recommendation algorithm organizes all this data and gives each category a specific weight. Netflix determines the weight of each category based on if it is deemed a more important contributing factor to predict another title that you will enjoy. For example, movies that have received universal acclaim like Citizen Kane or Schindler’s List still may not be the right match for many users, what is more important is what YOU have previously watched and enjoyed so this is given a greater weight. This algorithm will then group you into a taste preference. These taste preferences will have communities of users that enjoy the same type of content that you do. Based on what these users watch and enjoy in the future Netflix will also make recommendations for you.
Netflix employs “taggers” who watch all the content that Netflix has available and objectively tag them with characteristics that fit the movie. Through this Netflix has revolutionized the way that we see genre. Previously genres were limited to general descriptions with very few listed genres. Movies inside these genres may not have much in common even though they are in the same category. Netflix realizes that the reason for the user to watch a show changes based on the person. Todd Yellin, VP of Production of Innovation told the Associated Press a helpful way of understanding this: “We’ve found that people who tend to watch ‘Blacklist’ and ‘House of Cards’ tend to like ‘Ozark,’” Yellin said. “But another kind of person who will find he likes ‘Ozark’ is a fan of ‘Narcos’ and ‘El Chapo’ and other drug-cartel-oriented dramas and documentaries.”.
Netflix has an approximate 27,000 genres to sort movies into. These can range from “Cerebral” and “Workplace Comedy” to “Gritty British Prison Movies” and “Troubled Genius Dramas”. Netflix finds that classic genre constrictions are too limiting and to truly personalize recommendations you have to look past them.
Now that this data has been collected, organized, and the algorithm has determined the best recommendations for the user, the final step is how they will be presented to you. If Netflix finds great recommendations but doesn’t present them to you in a effective way then their efforts have been for nothing. The Netflix website has a layout where there are approximately 40 rows of titles, with each row holding movies that are recommended to you because they are in a similar category.
Everything about how the Netflix website looks to the user is calculated. Netflix estimates it takes 60–90 seconds for the consumer to lose interest if they cannot find anything that they would like to watch. This is why it is important that the first few rows have what Netflix estimates will be their best recommendations and what you will be most likely to watch. Their presentation comes through a copious amount of testing. A lot of this is called A/B testing, which is a pretty simple concept. Netflix will put two different versions of a presentation of the same thing and find out which has the most positive responses from the user.
All this effort put into Recommendations has paid off. Netflix estimates that they have saved 1 billion dollars per year in subscriber churn (amount of potential lost customers) through this recommendations system. While at first glance the process of recommending titles may seem simple to the user, Netflix puts in a lot of the work behind the scenes to find your next favorite Movie or TV show.