1. NETFLIX USE OF DATA SCIENCE: CASE STUDY

• Netflix wanted to improve movie recommendations to its audience by matching related movies to those they have watched and suggesting them to the audience. To achieve that goal, Netflix invested in data science to help push interesting and relevant film or TV show recommendations to the audience.
• After introducing an online streaming platform, Netflix invested in many algorithms to deliver a seamless movie experience to its audience. An example is the recommendation system to provide suggestions to the users by understanding their needs and providing suggestions on various films and TV shows.

Netflix’s Algorithmic Recommendation System

• Netflix’s recommendation system inputs user information, which could be in the form of past content consumed, product ratings, user-actions, etc., and processes it to predict user preferences for products and movie ratings. Such as recommendation system also deploys many machine learning algorithms.
• For cold start cases where Netflix lacks user past data information, it uses the recommendation system to search for similarity between different films or TV shows and provides recommendations of similar films or TV shows based on the movies being watched on the platform.
• Netflix recommendation system is either content-based or collaborative filtering based. For the former, Netflix similar suggestions to what a user watches; however, for the collaborative-filtering recommendation system, Netflix provides recommendations based on user profiles that are similar.
• The key data points taken into consideration for these recommendation systems include background information of its products and customer information. Other important data sets considered to push recommendations include the previously watched genres. The imagery below shows how the two systems work.
  Source
   

• Despite having the two recommendation systems, Netflix uses both of them via a hybrid recommendation system to suggest content to its users. Likewise, beyond the recommendation algorithms, Netflix also deploys "ranking algorithms" to provide ranked lists of movies and TV shows that appeal the most to its audience.
• With so many ranking algorithms in the market, Netflix implemented the "interleaving technique that allowed it to identify the best algorithms" that can deliver the best page ranking algorithm for improved personalized user recommendations. Moreover, to better predict contexts, Netflix uses representation learning, which leverages various user behaviors like "time and periods of watching, e.g., day, week, season, and even longer periods like Olympics, FIFA, and elections."
• Overall, the Netflix case study shows how it invested in data science to build algorithms that can make recommendations on movies or TV shows similar to those previously watched by its users, the many algorithms it uses to personalize user content, and the types of data sets and user actions it inputs for processing.

2. HULU USE OF DATA SCIENCE: CASE STUDY

• Hulu, just like Netflix uses many algorithms to power its recommendations feature for its audience. In 2017, the company started using Amazon Web Services (AWS) to support the addition of over 50 live channels for its Live TV offering.
• The media streaming giant opted for Amazon’s AWS to run its live TV services because AWS is reliable and secure enough to allow Hulu to deliver an unrivaled viewer experience, even when viewership or traffic spikes.
• Hulu is using different data science solutions to improve its platforms capability, functionality, and user experience. Among the technologies it deploys is the Video Genome Project it acquired in 2016 to improve recommendations for Live TV and VOD.

Personalization

• Hulu uses its advanced data capabilities and algorithms to personalize each user’s Hulu home screen and recommendations. Hulu aims to help its audience discover new shows based on a wealth of data that ensures the recommendations and personalization are accurate.
• The company relies on a combination of audience feedback, its advanced algorithms, and human curation to create the recommendations it personalizes for each user.
• The streaming platform includes a feature for users with options to like or dislike shows for submitting direct feedback to Hulu about the audience’s interests, to help the platform refine user results for improved viewership experiences.

Data and analytics.

• The Hulu platform has tied almost everything to data and analytics.
• The streaming platform deploys an "enhanced recommendation engine that tracks what users watch and when." The data and user action feedback acquired is then leveraged to develop a stronger and more personalized experiences.
• On Hulu, data drives everything regarding user experience, "from what a customer is recommended to the order the collections are displayed on the home screen."

Hulu’s Video Genome Project (VGP)

• Hulu’s VGP “dynamically aggregates metadata around video content.” The VGP has over 8 million records and can tag video component data automatically to define “genres on a granular level from basic sci-fi into alien or zombie.”
• The huge amount of data available on this platform helps the company to dynamically categorize video content into many genres. Hulu's VGP algorithms focus on everything, including less important ratings.
• Some notable data sets put into consideration include user actions, such as what film or TV show a user watched, searched for, rated, browsed, scrolled past, etc., combined with “time of day, device and geo-location, and larger audience trends.”
• Overall, research findings indicate that Hulu deploys data science to help it refine user viewership experiences, provide relevant recommendations to its audience, collect real-time feedback on the platform, etc., with the sole purpose of delivering a remarkable user recommendation and streaming experience.

Type of Data Used

• Data analytics in the film industry uses diverse data sets to provide real-time insights into the film's outcome. A review of modern AI driven solutions for predictive analytics in the film industry reveals the use of data sets like category/genre type, customer data, customer email addresses, unstructured data from social media platforms, data on people, data on content, etc.
• Common unstructured data utilized in AI in the film industry include text and imagery. Equally, specific consumer behavioral data, such as consumer history of film watching, including the types of films watched, consumers who consume film content, and those that do not consume it at all also form critical data sets that inform movie making decisions.
• Most of the data used in the film industry decision-making process comes from social media channels — Facebook, Twitter, Instagram, Snapchat, Google platforms, and YouTube, among others. Equally, the film industry can deploy programmatic analysis to help target individuals across different websites and sectors.
• A Los Angeles-based startup Cinelytic uses AI to analyze "historical data about movie performances over the years and cross-references the data with information about films’ themes and key talent, using machine learning to identify hidden patterns in the data." Its software allows users to even swap characters to see how it affects the anticipated performance of the movies.

How They Get the Data

• The film industry collects data but also relies on third-party vendors for both data and data analytics. For example, Hollywood relies on technology to convert movie-goer reactions content into insightful and meaningful data for decision-making.
• Hollywood is also working with IBM to deploy data and predictive analytics. IBM is helping Hollywood define the specific characteristics of audiences and tracking their sentiment across blogs, social media, film forums, etc.
• Online movie databases also provide a great source of data, on movie cast, character description, consumer feedback, movie ratings, etc. One such database is IMDB with more than half a million films, including datasets of movie dialogue subtitles.

How that Data Drives Business Decisions

• Data collected is helping movie executives make critical business decisions. For instance, by keeping close attention to the audience and the creativity, business executives are able to make informed decisions on how to make the movies, market, and distribute them.
• With data and predictive analytics, film executives can leverage real-time opinion assessments to flexibly target specific demographics in marketing campaigns, tailor trailers to appeal to individuals who prefer certain types of films, and pushing distribution to geographic locations where the target movie-goers live.
• Hollywood has for long relied on conventional methods of audience segmentation, i.e., the quad segment approach, however, working with IBM has brought new approaches, including the "profusion of very fine-sliced profiles that more accurately represents the sensibility spectrum of a modern audience." With such fine data, managers can leverage the feedback they receive from the profiled audiences to make better creative and marketing decisions.

• With very fine-sliced audience profiles, Hollywood can focus more on micro-segments; for instance, narrowing down to "soccer moms in Florida that are really passionate about action films," enabling film executives to access higher fidelity customer segmentation data that helps them understand the audience better for improved decision-making.
• By deploying data science in script assessment, film executives can "foster creativity and artistic expression by promoting counter-intuitive options that would be otherwise discarded by experts relying on common selection heuristics," considering that scriptwriters select scripts based on the best-selling novels, which can or cannot work well in theaters.