How Netflix uses Artificial Intelligence to Recommend Your Next Binge?
Understanding how Artificial Intelligence helps Netflix to Offer Personalized Experience for Every User
One weekend, you decided to binge-watch, ‘Strangers Things’ on Netflix. The next time you logged in to your Netflix account, you may notice recommendations based on the genre of ‘Strangers Things’ or any other series you have binged. Have you ever wondered how online streaming platforms like Netflix, offer you suggestions? Well, generally these platforms use a recommendation engine model of artificial intelligence for this. Let’s delve into how artificial intelligence (AI) helps in predicting your possible next watch.
Artificial intelligence refers to the simulation of human intelligence in machines, programmed to think like humans and imitate their actions. As an interdisciplinary field of computer science, artificial intelligence makes it possible for machines to learn from experience and perform human-like tasks effectively. Today, AI applications have never been more relevant, more important, or more complex.
Basics of Recommendation Engines
Recommendation engines can be considered similar to data filtering tools that use artificial intelligence algorithms and big data to recommend the most relevant items to a particular user. These tools can help online strea ming platforms like Spotify, Amazon Prime, Netflix, and e-commerce sites boost revenues, click-through rates, conversions, and more. While the recommendation engine is a part of artificial narrow intelligence (ANI) technology leveraged by Netflix, it is not the only use case. Basically, Netflix aims to employ both automation and human control to offer an enhanced personalized experience.
In today’s connected world, data is anointed as the fuel. With data explosion on the internet, and an increasing number of online users, companies must collate, map and analyze this data to offer services that align with their preferences. Typically recommendation engines follow a standard four-step process viz.,
1. Data is collected in two formats: Explicit data and Implicit data.
a) Explicit data includes feedback about the content given by users in the form of reviews and ratings, likes and dislikes, and product comments.
b) Implicit data is about collective user behavioral data like web search history, clicks, search and scrolling log, date, location and specific time one views content, device used, instances when users pause, rewind, fast forward, etc.
2. Data is stored based on its types and sybtypes.
3. Data analysis is carried out as per the requirement like real-time analysis, Batch analysis, near real-time analysis and more.
4. Data filtering is the final step where machine learning algorithms are applied to the data depending on whether collaborative, content-based, or hybrid model recommendation filtering is being used.
Collaborative filtering refers to analyzing data on user’s behaviors, activities, or preferences and predicting what they will like based on the similarity with other users. Content-based filtering focuses on the description and profile of the commodity. Here the algorithm uses data from one content source and replicates them across other different content types via the use of keywords, following a method called transfer learning. Unlike collaborative filtering, in this case, comparisons are done among the videos themselves based on a type of classification, like, for example, a genre.
And the hybrid model filtering is the combination of the other two filtering methods. Netflix uses hybrid model recommendation filtering.
According to Netflix research written by Carlos Uribe-Gomez and Neil Hunt published by the Association of Computing Machinery in 2015, the streaming site estimates the average user loses interest after 60 to 90 seconds if they have not found a worthwhile title to stream. The paper titled “The Netflix Recommender System: Algorithms, Business Value, and Innovation” mentions the following features used by Netflix to deliver a top-notch experience for its users:
· Personalized Video Ranker
· Top-N Video Ranker
· Video-Video Similarity
· Continue Watching
· Trending Now
· Page Generation: Row Selection and Ranking
· Statistical & Machine Learning Techniques for all of the above
According to an article on Thomasnet, genre rows on Netflix are determined by Personal Video Ranker. Within the genre rows, the Top N ranker finds the best matches for the user, using short-term trends. Video-to-Video similarity informs the viewer of options they might be interested in based on previously viewed titles.
Thumbnails in these suggestive videos play a crucial role in the user’s decision to watch the recommended titles. Each thumbnail has the potential to get users to spend their entire weekend watching Netflix shows and series, or unfortunately, force them to switch to a rival streaming platform like Disney+, Amazon Prime, HBO Go or Hulu.
Therefore, Netflix customizes the thumbnails of all its shows, as they are the first thing that provokes the user to watch a specific show. For every title, several thumbnail images are created randomly for different subscribers based on the user interest data. These thumbnails can either depict only the title, or portray the viewer recognized actors, the genre that the user prefers the most, a particular scene, ethnicity the user identifies with (or based on the user’s geographic location). Then, the images are annotated and ranked to predict the highest likelihood of being clicked by a user. These calculations are an aggregate of others with similar interests and preferences that have been clicked on.
Fine Tuning Streaming Quality
We all are aware of how important it is to have better streaming quality while watching anything online. A buffering, low-quality stream will imply frustration and switching off in no time. Artificial intelligence also helps Netflix review each frame of a video and compress it only to the degree necessary without degrading the image quality. Using the Dynamic Optimizer, the streaming giant, attempts to counter the bandwidth issues in emerging markets. Apart from enhancing streaming quality over slower speeds, this helps to customize content for customers that view Netflix on tablets and phones. Moreover, artificial intelligence algorithms help optimize audio and video encoding, in-house CDN, and adaptive bitrate selection.