Boasting its 209 million global subscribers (as of 2021 Q2), Netflix is no doubt the most popular video streaming platform. As its primary source of revenue comes from user subscription fees, Netflix focuses heavily on improving user retention rate through personalizing video recommendations. Its commitment to personalization is clearly expressed in this statement: “Personalization… enables us to not have just one Netflix product but hundreds of millions of products: one for each member profile.” Let’s take a look at how Netflix captivates users through offering personalized experiences.
Why is personalization crucial?
Capturing user attention is an important objective for video, social media and ecommerce platforms. The longer users spend time on the platform, the more advertisements they will see and the more likely they will purchase products. Personalized experience plays a key role in it by recommending content that matches users’ interests, thus keeping them on the platform longer. This is no easy task as humans have a very short attention span. Research suggests that the average Netflix user loses interest after 60 to 90 seconds of browsing. In other words, if Netflix is unable to entice a click from the user within this short amount of time, it is very likely that s/he will leave for another platform.
How personalized is Netflix’s recommendation system?
Netflix adopts various algorithms to recommend videos and provide relevant information through its Homepage, video details tab, emails and user notifications. On the Homepage, there are up to 40 rows of recommended videos grouped according to a common theme or category. As shown below, each row has a specific theme or category, e.g. TV Dramas Starring Women, Bingeworthy TV Dramas, Western TV Shows, etc.
Source: Netflix
● What themes or categories should be included in the homepage? (e.g. Is the user interested in Psychological TV dramas or Award-wining friendship dramas?)
● How to rank the themes?
● What videos should be included in each theme? How to rank the videos? (e.g. Should Aquaman or Fast and Furious rank at the top under the category Blockbuster movies?)
Netflix answers these questions through meticulous calculations by its recommendation algorithm, whose aim is to suggest videos that the user wants to watch. The organization of videos into rows of themes/categories is also a strategic move by Netflix. Not only does it make video selection convenient for users, it also allows Netflix to analyze user behavior and interests by looking at their scroll motion. When a user scrolls down, it indicates that s/he is not interested in the themes displayed on the screen; when the user scrolls left, it means that s/he is interested in the theme but not the top-ranking shows.
One movie, nine posters
Besides the Homepage, details of each show are also tailor-made for users:
● Which version of trailer and highlight should be played?
● How should the show be labelled?
● Which version of poster (Netflix calls them “artworks”) should be displayed to entice a click?
Netflix edited 10 versions of highlights for House of Cards to suit the interests of different segments of audiences. As for Stranger Things alone, there are 9 different potential artworks. Each artwork version highlights a unique feature of the film. It could be a close-up shot of a famous actor in the show, a classic scene, or one that best represents the film. The selection of artwork is based on past user behavior and interests. For instance, if you watched a lot action/adventure movies in the past, Netflix will use artworks that showcase action or exciting scenes when introducing new shows to you.
Source: netflixtechblog.com
Real-time collection of user behavioral data
Hyper-personalization would not be possible without a humongous amount of real-time data. Netflix’s recommendation algorithm constantly updates and modifies video recommendations based on different sets of user data, including:
● Video contents that were clicked on and searched by the user
● User interaction with the videos, including where a user pause, re-watch or leave and view duration, etc.
● User viewing habits, including the day, the time of day and device type
● Video contents that were not clicked by the user (i.e. videos that s/he is not interested in)
● Behavioral data of similar audiences
Every move on the platform by the user, no matter how insignificant it seems, is recorded and analyzed by Netflix. Also, user behavior that is more recent will weigh more in the calculation process.
When you click into a video, you can see a %-match score below the “Play” button. It is Netflix’s prediction on how much the video matches your interests.
Source: Netflix
Every user will receive a unique %-score for the same video. This is because users have different interests and viewing habits. A video may be a “91% match” for user A, but only a “71% match” for user B.
Improves system accuracy by encouraging user feedback
Besides collecting real-time behavioral data of users, Netflix encourages them to feedback on the relevancy of its recommendations. This helps its recommendation system to better understand each user’s interests and preferences. Users can click on the “Thumbs Up” (which means “I like this”) or “Thumbs Down” (which means “Not for me”) button on the video details tab. This tells Netflix whether you are interested the video, so that it can update the recommendation algorithm accordingly.
Source: Netflix
Netflix is extremely dedicated to delivering “the right content to the right person.” It is estimated that currently 80% of user watch time on Netflix comes from its recommendation algorithm. Data analysis plays a crucial role in Netflix’s success. It has also contributed to some of Netflix’s most successful purchases of worldwide hit series (e.g. House of Cards) in the past. We will share more on this in another article.
YouFind provides integrated digital marketing services including search engine optimization, online advertisements, social media, and more. We are passionate about learning and sharing the latest marketing news & strategies.
