How Does The Netflix Recommender System (NRS) Work?

NRS: COMBINATION OF CONTENT-BASED FILTERING & COLLABORATIVE FILTERING

The core feature of Netflix is the Netflix recommender System (NRS). NRS is algorithm-based operating system, and it is powered by a combination of content filtering based on the users’ past interactions with collaborative filtering algorithms based on a large database of user taste preferences (Pajkovic, 2021).

Filtering by content is based on a user’s previous data, which is gathered based on their interactions with the platform: viewing history, watch time, scrolling behaviour, etc. These data are combined with other large and intricate data sets that contain information derived from the 15,000 film and television titles offered by Netflix worldwide, including genre, category, actors, director, and release year, to produce recommendations and personalize a user’s experience (Wasko and Meehan, 2020). 

Collaborative filtering involves the same data extraction process but makes its recommendations according to a weighted combination of other users’ preferences, thus imitating person-to-person recommendations. Previously, the NRS’s collaborative filtering recommendations were limited to the data extracted from users in a specific region or country (Stenovec, 2016). Now recommendations are pulled from the viewing preferences of users across the world, and users themselves are algorithmically grouped into global “taste communities”.

Those titles that the NRS determines to be most relevant to a user land at the beginning of a row, and those rows appear higher on the homepage itself. Moreover, almost all information displayed regarding a specific title is personalized, including its match score, artwork, trailer, synopsis, and metadata (e.g. awards, cast, etc.).

CONTENT FLOW IN NETFLIX: “AUTOPLAY” AND “SEARCH”

Netflix prioritizes novel content, creating a sense of immediacy. In this regard, it greatly privileges its Autoplay function with two main goals; firstly, to feature and promote new releases; secondly, to encourage viewers to keep watching without their direct intervention (Johnson, 2019). How the Autoplay function works, and how it relates to content browsing and the partial marginalization of the Search function. Autoplay function works under the assumption that users will keep watching (McCormick, 2016; Stanfill, 2015). In other words, Netflix, as a media service, flows by default. As soon as the user opens its interface, without even choosing any content, Netflix automatically plays an excerpt of a featured product, which varies regularly depending on the company’s promotional strategies. Although users may act in a variety of different ways when using Netflix—letting Autoplay dictate what they watch, completely disregarding it, or occasionally use it—the Netflix interface is designed to require minimal work or decision making on the user’s behalf. 

Netflix’ software Cinematch is ‘a system that constantly translates seemingly chaotic behavior into recurring and therefore predictable patterns’ (Alexander 2016). Providing the illusion of never-ending choices, the Netflix interface organizes its thousands of titles into manageable categories through a complex set of recommendation filters. Users can avoid all these filters and utilize the ‘Search’ function, which is not centrally featured when one opens the Netflix interface but conveniently displaced to the top left side of the screen. If the searched content is available, Netflix does highlight the selected choice. However, if it is not, the search function activates a ‘similar titles’ logic, suggesting dozens of titles, attempting to satisfy users through a recommendation strategy that hides the lack of availability of a particular title. In other words, the Netflix Search function does not examine its catalogue but a vast audio-visual archive of potentially available audio-visual materials to then orient users toward what is available within a specific geographical territory, perpetuating the idea of endless choice. 

ILLUSION OF “PLENTY AND ABUNDANCE” & BROWSING

This is precisely one of the keys of Netflix’s success: the appearance of limitlessness, of making users believe the service has everything they may want, shrewdly hiding or downplaying the temporality of licensing agreements by constantly delivering new content, as a non-stop distribution and production machinery users need to constantly check if wanting to be in tune with the trending mediascape.

Browsing contributes to strengthen the misleading idea that the Netflix catalogue is limitless. Confronted with myriad choices and potentially attractive categories as they open the Netflix interface, users are encouraged to browse and browse, to seek a ‘better’ option even if a particular film or series may sound appealing. Facing this data overflow, users are both incredibly faithful to the series they cherish, the genres they prefer, or the latest special by their favorite comedian and, at the same time, (perhaps) naively hopeful that they will find their next beloved show as they keep browsing, hooked not only to the contents Netflix offers but also to the very dynamics of browsing. Netflix subscribers they do spend (and on occasion waste) hours contemplating the appearance of interminable choices the service offers, making this practice a fundamental part of their routine. 

⬇️ REFERENCES:

• Alexander, N. (2016). Catered to Your Future Self: Netflix’s “Predictive Personalization” and the Mathematization of Taste. In D. Smith-Rowsey (Eds.). The Netflix Effect: Technology and Entertainment in the 21st Century (pp. 81–98).
• Pajkovic N. Algorithms and taste-making: Exposing the Netflix Recommender System’s operational logics. Convergence. May 2021. doi: 10.1177/13548565211014464
• Rodríguez Ortega V. ‘We Pay to Buy Ourselves’: Netflix, Spectators & Streaming. Journal of Communication Inquiry. January 2022. doi: 10.1177/01968599211072446
• Wasko J and Meehan ER (2020) A Companion to Television. Hoboken, NJ: John Wiley & Sons.