Netflix Recommendation Algorithm

Before entering the digital library, Netflix customers choose their user journey by clicking on their individual profile icon.

Similar to social media, video streaming services are constantly vying for customers’ attention. Based on an advertising model, the three-steps goals of companies such as Facebook, Instagram, YouTube, and Netflix, can be condensed to engagement, growth, and revenue. Netflix deploys behavior science and personalization to drive user engagement.

“The best way to get someone’s attention is to know how their mind works.”

— Tristan Harris, TED Talk, 2017

To hook its customers, Netflix uses a recommendation algorithm called the Netflix Recommendation Algorithm (NNE), which is composed of multiple algorithms that filter content based on the users’ activity while engaged on the platform.

A guide on Netflix Recommendation Algorithm.

Infographic credit: Invisibly.

According to Invisibly, 80% of Netflix viewer activity is a direct result of personal recommendations and Netflix believes it could lose $1 billion or more from subscribers cancelling their service (2021).

Cognitive and Content Biases

Netflix deploys various content biases to keep its customers scrolling, as they search for the perfect tv show or film to watch. While a user may have intended to limit their watching to one 30-minute sitcom episode or conduct a quick library browse for a 90-minute movie, they are inundated with features, such as the system automated “play the next episode” or the seemingly endless video library. Consequently, the time elapsed while engaging in these activities often far exceeds the users’ predictions; however, Netflix succeeds in maintaining its customers’ attention for a long period of time, before the customer eventually disengages.

“It’s the gradual, slight, imperceptible change in your behavior and perception that is the product.”

— Jaron Lanier, The Social Dilemma (2020)

Today, social media and media streaming companies use algorithms to create unique, personalized UX experiences for each of their users rather than relying on a universal, static design. Conducting A/B tests enables Netflix to fine-tune experiences based on customers’ behaviors that indicate their preference. On Netflix’s part, this practice demonstrates confirmation bias, by delivering content that aligns with individuals’ preferences (Nivedha, 2024).

Users are presented with a library featuring over a thousand titles, Netflix deploys various tactics to keep their customers engaged and from not feeling overwhelmed in the process. Photo credit: Adobe Stock.

Netflix keeps its users engaged during the content search phase by creating an interactive browsing experience with personalized displays of preferred genres and the recently added feature of autoplay trailers when you hover over the title thumbnail. This interactive experience combined with intentionally making its original episodic series content between 30 to 45 minutes, Nivedha (2024) notes that the streaming service reduces cognitive load and diverts the potential for customer frustrations.

“Top-10 TV Shows in the U.S. Today” category. (November 17, 2024)

Another recently introduced feature on the platform, the carousel of “Top Ten” content, entices users to explore the most popular content in the users’ identified region. This allure is attributed to the Top-10 Effect. The phenomena is based on the idea that knowing others are watching a certain title would also increase the likelihood of us watching it, by way of social norms and social proof (Mobayed, 2024). By exploring a collection of culturally relevant content, users can feel reaffirmation of their social membership.

Even displayed content titles’ artwork is not static. The NRE deploys dynamic thumbnail artwork based on what the algorithm calculates as its users’ preference, and some have argued that Netflix uses content bias targeting based on the user’s race; however, 74% of customers feel frustrated when when website content is not personalized (Invisibly, 2021).

Analyzing My Netflix User Profile Recommendations

Examining my own Netflix user profile interface, my taste for a variety of genres and content in different languages are on display for me browse through. Before Netflix’s streaming platform, I often lost track of time browsing through the local video store or library’s collection; however, it is true that this past behavior was my choice and not largely attributed to machine learning manipulation.

Screenshot of my Netflix user profile’s Today Top Picks for You section (captured on November 26, 2024) showcases the first six out of 42 video recommendations based on algorithmic predictions calculated from the my past viewing behaviors.

Focusing on the algorithmically curated titles for my profile’s “Today’s Top Picks for You,” it is evident that my preference for one-hour episodic cooking/baking series populates the majority of the 42 thumbnail spots. The NRE has correctly predicted that recently, when I engage with the platform, I do not commit much time to longer length content.

Doing a quick review of the content presented to me, it is easy to write some content off as oddities. For example: from the thumbnail artwork alone, “Upper Middle Bogan”, the sixth suggestion that appears on the Top Picks screenshot above, does not look like content I would be interested in. Yet, after hovering over the title and watching the trailer, I recognized the tropes I usually gravitate towards, such as “quirky female leads in a 30-minute episode comedy series”

Watch Content on Netflix with Others

My significant other does not live nor shares a Netflix account with me. When this individual comes over to my residence and we decide to watch something on Netflix, it often becomes a frustrating experience for the both of us. Since we use my profile to access the library, I am content with the content curated for me, I allow my significant other to navigate with the remote and select the title. Yet, this experience consumes between 30 to 45 minutes of our time together and we are often irritated by the time a decision has been made. Perhaps establishing a joint profile, which is only used when we engage in Netflix together, will enable the NRE to output a UX interface where the combined preferences of my significant other and my self, will make for a more enjoyable time spent together. However, since Netflix’s goal is to keep its customers engaged for the longest periods of time, I doubt we would spend any less time deciding on a title to watch.

---

References

Harris, T. (2017, April). How a handful of tech companies control billions of minds every day [Video]. TED Conferences. https://www.ted.com/talks/tristan_harris_how_a_handful_of_tech_companies_control_billions_of_minds_every_day?subtitle=en

Invisibly. (2021, November 10). Behind The Scenes of The Netflix Recommendation Algorithm. https://www.invisibly.com/learn-blog/netflix-recommendation-algorithm/

Mobayed, T. (2024, August 8). Netflix, Behavioral Science, and Personalization. Psychology Today. https://www.psychologytoday.com/us/blog/emotional-behavior-behavioral-emotions/202408/netflix-behavioral-science-and-personalization

Nivedha. UX Magic: How Netflix Uses Cognitive Biases to Keep Us Hooked. Medium. https://medium.com/design-bootcamp/ux-magic-how-netflix-uses-cognitive-biases-to-keep-us-hooked-b9f40b3f8efe

Orlowski-Yang, J. (Director). (2020). The Social Dilemma [Film]. Exposure Labs; Argent Pictures.