Hulu recently refreshed its mobile app with an update that provides its users with more control over the way the recommended shows are presented. Users are now able to tell Hulu to stop suggesting content they have no interest in watching, removing the series, movie or sports league from being recommended again. Moreover, they are able to remove items from their viewing history and Hulu’s recommendation engine will essentially forget that the user ever watched the show or movie in question.
Another streaming service that needs no introduction is Netflix, with an offer of thousands of TV shows and movies available for streaming and download. Netflix recommends titles for each user depending on their viewing and browsing history and no users have the same recommendations.
To classify its suggestions, Netflix creates precise genres (i.e.“Crime thriller where the main character is female”).
- How are those genres created?
- >How does Netflix manage to provide meaningful recommendations to its 100 million-plus subscribers who are already getting recommendations from every platform they use?
The answer is Machine learning, algorithms, and creativity.
The recommendation system works putting together data collected from different places.
Recommended rows - are tailored to the users viewing habits. Systems based on machine learning continuously rewrite themselves as they learn from their own users. Every time a user spends some time watching a TV show or a movie, Netflix is collecting data that informs the algorithm and refreshes it. The more you watch the more precise the algorithm is.
The tags used for the Netflix machine learning algorithms are the same across the globe. In fact, real-life humans were hired from the company to categorize every TV shows and movies offered by the streaming service and apply tags to each of them in order to create specific micro-genres such as “Critically acclaimed road trip movies”. Each movie watched by a particular user is identified in a particular taste group. Each user’s screen gets populated (left, right, and top to bottom) based on which groups they belong to.
The main challenge encountered by Netflix consists of creating useful groupings of movies or tv-shows and highlighting the variety and the depth of the offer, helping users not only to reinforce their areas of interest but also to find new ones. The recommendations have to be fresh but also stable so that people that are used to their homepage can easily find videos they’ve been recommended in the recent past.
Netflix - Looking at the future
A UX study on Netflix users from late 2018 included two major findings:
- According to users, one of the major pain points when browsing the content is not finding relevant content fast enough. Users are spending a lot of time scrolling through irrelevant content that they won’t engage with.
- Users don’t consume the same content every time they log in. In fact, the content differs according to parameters such as time, day of the week, location, device, etc.
Because of these findings, Netflix’s object is to implement a recommendation system that knows which content is the most relevant for the specific user in relation to the time and parameters of the login. Netflix is working on a new discovery experience to know its users better and provide them with recommendations that take into account different parameters. The goal of this experience is to recommend content that users are most likely to engage with at that specific time of the day on a particular device.
As an avid Netflix user, it’s very true that the shows and movies that I want to watch in the morning on weekends are different from the ones that I want to watch while commuting or at night. Looking at the future, Netflix’s progress is tied to its ability to innovate and find a way to integrate other parameters such as time, location, day of the week and device in its algorithms.