• Computer architecture design has evolved towards adding various levels of caching between compute units and the main memory to hide latency. These caches are partially shared among the CPUs, so perfect performance isolation of co-hosted containers is not possible. Netflix's team was able to meaningfully improve both the predictability and performance of its containers by taking some of the CPU isolation responsibility away from the operating system and moving towards a data-driven solution involving combinatorial optimization and machine learning. This post explains the process in detail.

    Friday, April 12, 2024
  • Netflix heavily uses Java for its backend apps. It started with a microservice architecture that evolved over time to incorporate new technologies. Initially, the company employed Groovy scripts and reactive programming for tailored API responses. It has since migrated to GraphQL Federation for a unified API, removing the need for custom backend layers. Netflix uses recent Java versions for performance gains and actively uses Spring Boot.

    Thursday, April 4, 2024
  • Netflix's Graph Search has improved to include reverse search functionality, enabling precise notifications and dynamic updates for users. By using Elasticsearch and CockroachDB, Netflix efficiently handles SavedSearches, improving performance. The system's flexibility also allows for expanding reverse search to other teams and exploring responsive UIs through subscriptions.

  • Netflix's latest earnings report showcases its remarkable ability to attract and retain subscribers. The company faced several challenges such as password-sharing crackdowns and pricing adjustments, but was able to successfully adapt to changing market dynamics and maintain its leadership position. Netflix focused on delivering value through a combination of compelling content, personalized recommendations, and strategic initiatives like live events and gaming.

  • Apple will soon begin basing pay on how a series or movie performs. Talent would receive bonuses on a points system, with the size of the bonuses being based on the number of people who signed up for Apple TV+ to watch, how much time they spent viewing, and the cost of the program relative to the size of its audience. Netflix and Amazon have also spent months developing plans for performance-based compensation, but neither have gone public yet with their models. Streaming services are becoming increasingly more transparent - the practice will make it easier to tie pay to performance.

  • Amazon and Netflix have signed a three-year deal that will put exclusive Christmas games on the streaming service. The first two games will be aired on Christmas Day. 2025 and 2026 will see at least one game on the service each Christmas. The NFL is able to cancel all of its TV deals at the same time in 2029 - this could lead to an unprecedented bidding war that could upend the entire NFL content world.

  • Netflix's ad-supported plan now has 40 million monthly active users, with more than 40% of all signups in ads markets picking the plan. The Trade Desk, Google's Display & Video 360, and Magnite will join Microsoft this summer as Netflix's main programmatic partners for advertisers. Netflix has announced plans to launch an in-house ad tech platform. It also announced a slew of new content, including two NFL Christmas games, and a slate of new measurement partners.

  • Netflix's introduction of ad-supported subscription tiers marks a significant shift in its business strategy that aims to appeal to price-sensitive consumers and diversify its revenue. By partnering with advertising firms and offering cheaper plans, Netflix hopes to attract a broader audience and compete with other streaming services. Netflix plans to innovate in advertising approaches, develop strategic partnerships, and ensure that ad-supported models contribute to long-term sustainability and growth.

  • Netflix is testing a homepage redesign on its TV app that replaces the static tiles with boxes that extend when they're selected. It is also removing the menu that pops out from the left side of its homepage and replacing it with a more streamlined selection of options at the top of the screen. The new homepage will be tested among a small group of users to start and expanded to more members if it goes well.

  • Netflix's first major redesign in a decade replaces the left-hand menu with a horizontal menu bar. Users can now hover over individual tiles for a beat to trigger a clip from a show or movie along with a text description and more information about it. The "New and Popular" and "My List" tabs are gone, the "Discover" feature has been made more prominent, and navigation has been simplified to adapt to a new streaming landscape that includes live sports and games. Netflix is testing the latest app version with a few subscribers before making it available.

  • TV networks are struggling in the current upfront market as advertisers reduce their spending and demand significant rate rollbacks. With an oversupply of streaming ad inventory and reduced demand, networks like Disney, NBCUniversal, and Fox are aggressively negotiating deals, while Amazon and Netflix hold out for higher rates. The industry also faces challenges from a lack of new content due to last year's Hollywood labor strike and cautious advertisers, further complicating the market.

  • Netflix plans to open two new in-person experience venues in 2025. The locations, which will be in King of Prussia, Pa., and Dallas, will feature a wide array of shopping outlets, eateries, and activities tied to major franchises. The two new Netflix Houses will have footprints spanning more than 100,000 square feet. Netflix doesn't see these permanent retail destinations as becoming a sizable new business segment - it is aiming for them to serve as marketing vehicles that invite fan engagement to support the core subscription-streaming business.

  • While Netflix doesn't release a lot of its numbers to the outside world, the streaming giant has long been aggressively transparent internally. Netflix trusts its employees to act in the best interest of the company. Employees are encouraged to speak about what's going well and what's not on a regular basis. Netflix has a culture of trying to systematically think about what generates long-term excellence. Many more details about Netflix's internal culture are available in the article.

  • Although Netflix leads in streaming, it faces a tough challenge in advertising. It is only the ninth or tenth largest player. The company's efforts to expand its ad-supported tier have seen mixed results, with under-delivered impressions and strained partnerships. Netflix is exploring live events and sports to increase its advertising inventory and reach and compete with industry leaders like Amazon and Disney.

  • Netflix initially implemented prioritized load shedding at the API gateway level, then extended it to individual service level for finer-grained control and better cloud capacity utilization. One example is its PlayAPI, which prioritized user-initiated requests over pre-fetch requests during an infrastructure outage, maintaining high availability for critical requests. Netflix also created an internal library for prioritized load shedding and experimented with CPU-based and IO-based load shedding.

  • Netflix is set to open Netflix House, an immersive retail and entertainment venue based on its hit series, at King of Prussia Mall in Pennsylvania and Galleria Dallas in Texas. The venue will offer experiences that feature sets, games, restaurants, and shops inspired by shows like Bridgerton and Squid Game. The location will span over 100,000 square feet and the experiences will rotate year-round. Opening in 2025, the concept aims to bring Netflix's beloved stories to life in new and ever-changing ways.

  • Netflix's growth model is evolving with the rise of ad-supported memberships, which grew 34% quarter-over-quarter and now account for 45% of sign-ups in available markets. Despite this rapid growth, Netflix's current ad inventory can't keep pace, so significant ad revenue isn't expected until after 2025. To address this, Netflix is developing an in-house ad tech platform set to launch globally in 2025.

  • Runway's AI text-to-video generator was trained using thousands of YouTube videos and pirated films sourced from significant entertainment channels and news outlets. The broad dataset includes videos from Netflix, Disney, and creators like MKBHD and Linus Tech Tips. YouTube's CEO noted that using its videos for AI training is a policy violation, but other prominent tech firms have also been linked to similar practices.

  • Netflix is launching Netflix House, an experiential retail activation, to revive mall foot traffic and engage consumers with its programming. The first locations will open next year in Pennsylvania and Texas malls, offering immersive experiences, retail therapy, and themed food and drink based on popular Netflix series. Netflix hopes to build deeper connections with viewers by providing unique, updated experiences tied to franchises like Stranger Things and Bridgerton.

  • Netflix's video game strategy is expanding, especially for games tied to its reality TV shows. Third-party data indicates success, with popular mobile games achieving 68 million downloads, boosted by the Grand Theft Auto franchise. Netflix focuses on tie-in games to engage superfans between show seasons, showing the potential of cross-platform marketing. The company is testing cloud gaming, allowing more complex games on smart TVs, and integrating marketing for mobile games into its TV app.

    Hi Impact
  • A popular tactic currently is "Mini-Games Marketing," where brands use simple games to capture and retain user attention. Inspired by Toggl's "Startup Simulator," companies like LinkedIn, YouTube, and Netflix have adopted mini-games to increase interaction. These games work because they keep users on-site longer, foster brand loyalty (by making users feel understood), and have high viral potential, as players share scores and challenges. Mini-games also stand out by being a more memorable experience compared to traditional content.

  • Netflix researchers developed methods to analyze historical A/B test data to understand the relationship between proxy metrics and long-term business outcomes. They propose three estimators that offer more robust solutions for understanding the true relationship between proxy metrics and north star metrics: Total Covariance (TC), Jackknife Instrumental Variables Estimation (JIVE), and Limited Information Maximum Likelihood (LIML).

  • Netflix's Key-Value Data Abstraction Layer (KV DAL) addresses challenges the company had with datastore misuse by providing a consistent interface layer over storage to application developers. This abstraction offers a two-level map architecture and supports usage through basic CRUD APIs, complex multi-item and multi-record mutations, and efficient handling of large blobs through chunking. It uses idempotency tokens, client-side compression, and adaptive pagination for predictable performance.