Netflix

== Technology ==
=== Content delivery ===
Netflix freely [[Peering|peers]] with [[Internet service provider]]s (ISPs) directly and at common [[Internet exchange point]]s. In June 2012, a custom [[content delivery network]], [[Open Connect]], was announced.<ref>{{Cite web |author=Ryan Lawler |date=June 4, 2016 |title=Netflix Rolls Out Its Own CDN: Open Connect |url=https://techcrunch.com/2012/06/04/netflix-open-connect/ |access-date=September 12, 2016 |work=Tech Crunch}}</ref> For larger ISPs with over 100,000 subscribers, Netflix offers free ''Netflix Open Connect'' [[server appliance|Computer appliance]]s that cache their content within the ISPs' [[data center]]s or networks to further reduce [[Internet transit]] costs.<ref>{{cite web |title=Netflix Open Connect Content Delivery Network |url=https://www.netflix.com/openconnect |access-date=October 25, 2014 |work=netflix.com}}</ref><ref>{{cite web |author=Eric Savitz |date=June 5, 2012 |title=Netflix Shifts Traffic To Its Own CDN; Akamai, Limelight Shrs Hit |url=https://www.forbes.com/sites/ericsavitz/2012/06/05/netflix-shifts-traffic-to-its-own-cdn-akamai-limelight-shrs-hit/ |access-date=October 25, 2014 |work=Forbes}}</ref> By August 2016, Netflix closed its last physical data center, but continued to develop its Open Connect technology.<ref>{{Cite news |author=Peter Judge |date=August 20, 2015 |title=Netflix's data centers are dead, long live the CDN! |work=Data Center Dynamics |url=http://www.datacenterdynamics.com/content-tracks/colo-cloud/netflixs-data-centers-are-dead-long-live-the-cdn/94661.fullarticle |access-date=September 15, 2016}}</ref> A 2016 study at the [[University of London]] detected 233 individual Open Connect locations on over six continents, with the largest amount of traffic in the US, followed by Mexico.<ref>{{Cite news |author=Richard Chirgwin |date=June 22, 2016 |title=Boffins map Netflix's Open Connect CDN: Six continents, 233 locations, thousands of servers |work=The Register |url=https://www.theregister.co.uk/2016/06/22/boffins_map_netflixs_open_connect_cdn/ |access-date=September 15, 2016}}</ref><ref>{{Cite journal |last1=Böttger |first1=Timm |last2=Cuadrado |first2=Felix |last3=Tyson |first3=Gareth |last4=Castro |first4=Ignacio |last5=Uhlig |first5=Steve |date=January 2018 |title=Open Connect Everywhere: A Glimpse at the Internet Ecosystem through the Lens of the Netflix CDN |journal=ACM SIGCOMM Computer Communication Review |volume=48 |issue=1 |arxiv=1606.05519 |bibcode=2016arXiv160605519B |doi=10.1145/3211852.3211857 |orig-year=Submitted June 17, 2017 |s2cid=215824680}}</ref>

As of July 2017, Netflix series and movies accounted for more than a third of all prime-time download Internet traffic in North America.<ref>{{cite news |last=Ng |first=David |date=July 29, 2017 |title=Netflix is on the hook for $20&nbsp;billion. Can it keep spending its way to success? |work=[[Los Angeles Times]] |url=https://www.latimes.com/business/hollywood/la-fi-ct-netflix-debt-spending-20170729-story,amp.html |url-status=live |url-access=limited |archive-url=https://web.archive.org/web/20171023223520/http://www.latimes.com/business/hollywood/la-fi-ct-netflix-debt-spending-20170729-story,amp.html |archive-date=October 23, 2017}}</ref>

=== API ===
On October 1, 2008, Netflix offered access to its service via a public [[API|application programming interface]] (API).<ref>{{cite web |title=Netflix API Launches Tomorrow |url=http://www.readwriteweb.com/archives/netflix_api_launches_tomorrow.php |url-status=dead |archive-url=https://web.archive.org/web/20081002035838/http://www.readwriteweb.com/archives/netflix_api_launches_tomorrow.php |archive-date=October 2, 2008 |access-date=September 30, 2008 |publisher=ReadWriteWeb}}</ref> It allowed access to data for all Netflix titles, and allows users to manage their movie queues. The API was free and allowed commercial use.<ref>{{cite web |title=Netflix API Management Solution |url=http://www.mashery.com/customers/netflix |url-status=dead |archive-url=https://web.archive.org/web/20130216123114/http://www.mashery.com/customers/netflix |archive-date=February 16, 2013}}</ref> In June 2012, Netflix began to restrict the availability of its public API.<ref>{{cite web |date=June 15, 2012 |title=Upcoming Changes to the Netflix API Program |url=http://developer.netflix.com/blog/read/Upcoming_Changes_to_the_Netflix_API_Program |url-status=dead |archive-url=https://web.archive.org/web/20120618004836/http://developer.netflix.com/blog/read/Upcoming_Changes_to_the_Netflix_API_Program |archive-date=June 18, 2012 |access-date=June 21, 2012 |publisher=Netflix}}</ref> Netflix instead focused on a small number of known partners using private interfaces, since most traffic came from those private interfaces.<ref>{{Cite web |author=Daniel Jacobson and Sangeeta Narayanan |date=July 24, 2014 |title=Netflix API: Top 10 Lessons Learned (so far) |url=http://www.slideshare.net/danieljacobson/top-10-lessons-learned-from-the-netflix-api-oscon-2014 |access-date=September 12, 2016}}</ref> In June 2014, Netflix announced it would be retiring the public API; it became effective November 14, 2014.<ref>{{Cite web |author=Janko Roettgers |date=November 14, 2014 |title=Netflix is shutting down its public API today |url=https://gigaom.com/2014/11/14/netflix-is-shutting-down-its-public-api-today/ |access-date=September 12, 2016 |work=GigaOm |archive-date=November 15, 2016 |archive-url=https://web.archive.org/web/20161115211830/https://gigaom.com/2014/11/14/netflix-is-shutting-down-its-public-api-today/ |url-status=dead}}</ref> Netflix then partnered with the developers of eight services deemed the most valuable, including Instant Watcher, Fanhattan, Yidio and Nextguide.<ref>{{cite web |last=Jacobson |first=Daniel |date=June 13, 2014 |title=Retiring the Netflix Public API |url=http://developer.netflix.com/blog/read/Retiring_the_Netflix_Public_API |url-status=dead |archive-url=https://web.archive.org/web/20140614024656/http://developer.netflix.com/blog/read/Retiring_the_Netflix_Public_API |archive-date=June 14, 2014 |access-date=August 11, 2014 |publisher=Netflix}}</ref>

=== Recommendations and thumbnails ===

Netflix presents viewers with recommendations based on previous viewing history and ratings of viewed content. These are often grouped into genres and formats, or feature the platform's highest-rated content. Each title is presented with a thumbnail. Before around 2015, these were the same [[key art]] for everyone, but since then has been customized. Netflix may select a specific actor for a thumbnail based on viewing history,<ref>{{cite web |url=https://www.looper.com/274997/the-secret-behind-netflixs-personalized-thumbnails/ |title=The Secret Behind Netflix's Personalized Thumbnails |author=Dany Roth |date=November 8, 2020}}</ref> or an actor or scene type based on genre preferences.<ref>{{cite web |url=https://netflixtechblog.com/artwork-personalization-c589f074ad76 |title=Artwork Personalization at Netflix |work=Netflix Technology Blog |date=Dec 7, 2017 |author1=Ashok Chandrashekar |author2=Fernando Amat |author3=Justin Basilico |author4=Tony Jebara}}</ref> Some thumbnails are generated from video stills.<ref>[https://blogs.cornell.edu/info2040/2022/09/28/how-netflix-uses-matching-to-pick-the-best-thumbnail-for-you/ How Netflix Uses Matching To Pick The Best Thumbnail For You]</ref>

The Netflix recommendation system is a vital part of the streaming platform's success, enabling personalized content suggestions for over 220 million subscribers worldwide.<ref name="RecoAI">{{cite web |title=Netflix Recommendation System: How it Works |url=https://recoai.net/netflix-recommendation-system-how-it-works/ |work=RecoAI |date=April 5, 2022 |access-date=March 28, 2024}}</ref> Using advanced machine learning algorithms, Netflix analyzes user interactions, including viewing history, searches, and ratings, to deliver personalized recommendations for movies and TV shows.

The recommendation system considers individual user preferences, similarities with other users with comparable tastes, specific title attributes (genre, release year), device usage patterns, and viewing time. As users interact with the platform and provide feedback with their viewing habits, the recommendation system is able to adapt and refine its suggestions over time. Netflix uses a two-tiered ranking system, using the presentation of titles on the homepage for easy navigation to maximize user engagement. This is done by organizing content into rows and ranking the titles within each row based on how much the user would be interested in it.<ref name="RecoAI"/> Netflix also uses A/B testing to determine what causes the biggest interest and engagement related to options concerning movie suggestions and how titles are organized.

Tags like "bittersweet", "sitcom", or "intimate" are assigned to each title by Netflix employees.<ref name="Koblin" /> Netflix also uses the tags to create recommendation micro-genres like "Goofy TV Shows" or "Girls Night In".<ref name="Koblin">{{cite news |newspaper=[[The New York Times]] |author=John Koblin |date=January 14, 2024 |title=A Few Words About Netflix's Success: Vivid. Snappy. Tags. |url=https://www.nytimes.com/2024/01/14/business/media/netflix-streaming-movies-ratings.html}}</ref>