NetFlix using Machine Learning to maximize customer experience: A Case Study

What began as a DVD rental service in 1998 is now one of the world’s most powerful and renowned media streaming services. Having garnered subscribers up to almost 158.3 million, an estimate of nearly 37% of the world’s internet users are using Netflix.

Having Netflix has become the “in thing” among today’s populace and “binge-watching”, a formerly alien term has now become almost interchangeable with this service. Be it cartoons, movies, original web series, TV series, and documentaries available in multiple languages as well as subtitles and varied genres and categories, Netflix has something to offer to all generations and majority nationalities.

Use of Data Science and algorithms to improve customer experience

In the early 2000s, Netflix had initiated an open competition offering a 1-million-dollar prize for the best collective filtering algorithm to predict the ratings of users for films, based on their previous ratings. This approach resulted in becoming the turning point for the service.

Now, Netflix uses an opulence of technological algorithms to boost and enhance its customer experience.

Below are a few approaches using Data Science which are adopted by Netflix to improve the customer experience:

1. Recommendation System

In a service like Netflix, every action the user takes is recorded. The shows watched, the time of day when they are viewed, what was watched before and after that show, how quickly a series is binge-watched, when and where a user gets bored and stops watching, how long does a user take to scroll, and every single click of the pause and play button. Using a detailed tagging system Netflix can recommend its users the content which it knows will be their cup of tea.

Recommendation Systems are mainly of two main types:

• Content-based Recommendation Systems: In this system, the background knowledge regarding the products as well as the customer information is taken into account. Similar suggestions are provided based on the content the user has viewed on Netflix. For example, if the user has watched a film that has a “thriller” genre, similar films, having the same genre will be suggested.
• Collaborative filtering Recommendation Systems: This system provides suggestions based on the similar profiles of its users and is independent of knowledge of the product. This system is based solely on the assumption that what the users prefer in the past they will also prefer in the future.

2. Personalised Thumbnails

One thing which can be noticed upon opening Netflix is that the thumbnail you will see for a particular movie or show may not be similar to the thumbnail another user will get. Netflix elucidates the thumbnail images and then ranks every image in an attempt to gauge which thumbnail will have the maximum possibility of getting clicked by a particular user. These calculations are mainly based on what users similar to that particular user have clicked on.

One discovery can be that users who like a certain actor or movie genre have a greater likelihood of clicking images with that certain actor or image depicting a certain scenario.

3. Optimized streaming quality

Netflix makes use of past viewed data for predicting bandwidth usage to help the service decide when it should cache regional servers to ensure prompt load times during high demand. Thus, the service predicts which show is to be streamed in a certain location and caches the content in the nearby server when the internet traffic is minimal. This is done to ensure that the content is streamed without any buffering to maximize customer satisfaction.

Netflix undergoes a rigorous process of A/B testing for its adaptive streaming as well as content delivery network algorithms before it becomes the default customer experience.

4. Finding the next hit series

Netflix spent around 100 million on 26 episodes of House of Cards by analyzing the data of its viewers, i.e. based on their customer behavior as well as feedback. Netflix knew through its use of Big Data that its customers liked Kevin Spacy and through correlation, they derived that their users also liked David Fincher.

Hence with this data, Netflix was able to bring together the ideal cast as well as director for the show. Even the show’s poster was generated through machine learning.

Conclusion

Predominantly the reason for Netflix’s growing popularity is its programmed intuitive nature. From knowing what content to offer to which user, creating a user’s profile to provide personalized services to each of them as well as enhancing their streaming experience and offering a variety of such convenient features, this article highlights the way the service has made exceptional use of Data handling and analytics to reach where it is today and to maximize its customer’s experience as well as to promote its brand.