How Netflix Reverse-Engineered Hollywood

by Alexis C. Madrigal Read the Original

Netflix utilizes large teams of people to watch films and tag them with metadata, along with millions of users' viewing habits, to create a sophisticated database of 76,897 unique ways to describe types of movies. This data, combined with "Netflix Quantum Theory" and a 24-page document outlining how to tag movies, is used to create personalized genres and recommendations for users. Additionally, a strange Perry Mason bug is a result of combining human and machine intelligence.

Netflix has created a database of 76,897 unique ways to describe types of movies, based on large teams of people who watch films and tag them with metadata. This data, combined with millions of user's viewing habits, is a key part of their strategy to gain and retain subscribers. “Members connect with these [genre] rows so well that we measure an increase in member retention by placing the most tailored rows higher on the page instead of lower,” Personalization is key for customer loyalty. Netflix has found success in tailoring its recommendations to users, leading to higher retention rates.

Netflix has an absurdly large number of genres, more than previously thought. A bot was created to scrape all the data and revealed patterns in the data, but note that the existence of a genre doesn't necessarily mean that movies are available to stream. "Netflix had an absurdly large number of genres, an order of magnitude or two more than I had thought" Netflix has an incredibly large catalog of genres, most of which are unknown to the average user. This helps to personalize movie recommendations and create an ever-expanding library of content.

Madrigal studied Netflix's grammar to build a generator that decodes their vocabulary and categories, such as adjectives, countries of origin, nouns, time periods, subjects and age-specific genres. The generator outputs amazing genres, like "Post-Apocalyptic Comedies About Friendship". "The single-word adjectives (such as romantic) could basically just pile up" Netflix uses a very structured approach to genre formation, which allows them to create complex genres from simple components.

Netflix VP of Product Todd Yellin created a 24-page document outlining how to tag movies and guide the creation of all their systems. Madrigal tracked his intelligence and sensibility through the data, and when I visited the company's Silicon Valley office I found out he had been waiting for someone to discover it. "As we worked on the generator, I could tell someone had gone down this road before. A single human brain had had to make the decisions that we had. How many adjectives? How long should they be? And even more basic: what should the adjectives be? Why cerebral and not brainy?" Netflix has a sophisticated system to tag movies with specific adjectives and genre descriptions, indicating a lot of thought and research has gone into creating the system.

Netflix developed "Netflix Quantum Theory" to systematically tag thousands of movies into microtags such as "social acceptability" and "romance rating". This data is used to create personalized genres which provide an introspective tool for users to discover movies they may like. “My first goal was: tear apart content!” Netflix recognized the need to create a comprehensive system for analyzing the content within their movies. This was the first step in understanding the data and creating their personalized genres.

Netflix has a hybrid human and machine intelligence approach to personalizing movie recommendations. It uses an algorithm to predict how users rate movies, and a 36-page tagging document to understand the content of each movie. As an example, they might recommend an action movie with the right amount of romance for the user. This system has no analog in the tech world, making it even more impressive. “It is, in its own weird way, a tool for introspection.” Netflix's personalized genres provide users with a means of introspection. Through its analysis of movie data, Netflix can provide users with detailed descriptions of movies that offer insight into the user's preferences.

Netflix's actor-based genre-creation doesn't make much sense, except for a weird Perry Mason thing. It appears Raymond Burr and Barbara Hale are featured prominently due to fans ordering a lot of their movies in the DVD days. This strange occurrence is a result of combining human and machine intelligence and is sometimes called a bug and sometimes called a feature. “We wanted to highlight our personalization because we pride ourselves on putting the right title in front of the right person at the right time.” Netflix takes great pride in tailoring their movie recommendations to the individual user. They understand that providing the perfect movie for each user increases their chances of watching and enjoying it.