KE1251

t
January 27, 2023

os
rP
KEVIN MCTIGUE AND THEO ANDERSON

yo
Disney+ and Machine Learning
in the Streaming Age
Did Hopper die?
op
That was one of many questions Margaret Gupta wrestled with as she went to bed. It was July
2019, and Gupta had just finished the season-three finale of the Netflix show Stranger Things.
Although she had planned to watch just the sixth episode, she ended up bingeing the last two as
well, and now it was late. She sighed and looked at her watch, thinking that she might well prefer
meeting the show’s monster Demogorgon in a shopping mall to the business meeting she had in
tC

three days.

Gupta, a senior data scientist at the Walt Disney Company, was working on Disney+, the
subscription video-on-demand service that Disney planned to launch in November as a rival to
Netflix. Senior management was particularly interested in how the data-science team could help
the company navigate the uncharted waters of the streaming era, and the many new competitors
and challenges it presented. Netflix, for example, had pioneered the use of machine learning (ML)
No

to optimize its content, recommendations, and streaming quality. Superior ML helped Netflix
grow from a startup in 1997 to the leading streaming service in 2019, with more than 60 million
customers in the United States. The second-place service, Amazon Prime Video, had about 41
million subscribers.1 Disney+ would also compete for subscribers with Apple, which planned to
launch its own video-on-demand service in late 2019.

It was a daunting challenge: Disney’s competitors were some of the world’s most innovative
and resource-rich tech companies. Disney’s senior management wanted Gupta to offer ideas for
Do

© 2023 by the Kellogg School of Management at Northwestern University. This case was prepared by Professor Kevin
McTigue and Theo Anderson. Cases are developed solely as the basis for class discussion and are not intended to
serve as endorsements, sources of primary data, or illustrations of effective or ineffective management. This case was
based on publicly available information. For pedagogical purposes, the authors might have fictionalized individuals,
conversations, strategies, assessments, or other details. To order copies or to request permission to reproduce materials,
call 847.491.5400, or email [email protected]. No part of this publication may be reproduced, stored
in a retrieval system, used in a spreadsheet, or transmitted in any form or by any means—electronic, mechanical,
photocopying, recording, or otherwise—without the permission of Kellogg Case Publishing.

This document is authorized for educator review use only by ARPITA SRIVASTAVA, GL Bajaj Institute of Management & Research (GLBIMR) until Aug 2023. Copying or posting is an
infringement of copyright. [email protected] or 617.783.7860
 Disney+ and Machine Learning KE1251

t
how ML could help Disney and Disney+ not only survive in this environment but thrive. She

os
would present her ideas about the most valuable use cases in just three days.

Disney’s Origins, Reinventions, and Expansions

Founded in Hollywood, California, in 1923 as a film studio that made short cartoons, Disney

rP
by the middle of the 20th century had become a major force in the movie and live-entertainment
industries. In 1937, it released Snow White and the Seven Dwarfs, the first full-length, animated
feature in full color and sound. In 1954, it entered the television market with a weekly anthology
show, Walt Disney’s Disneyland, on the ABC television network. That program helped promote
its first theme park, Disneyland, which opened in the Los Angeles suburb of Anaheim in July
1955. Through the 1950s and early 1960s, a golden era of Disney animated movies—including

yo
Alice in Wonderland (1951), Peter Pan (1953), Sleeping Beauty (1959), and One Hundred and
One Dalmatians (1961)—cemented Disney’s reputation and created content for the company’s
television and theme-park endeavors. (See Exhibit 1.) The founder, Walt Disney, died in December
1966, just after construction began on Disney’s second theme park, Walt Disney World, which
opened near Orlando, Florida, in 1971.

Disney’s output and creativity declined through the 1970s, and by the early 1980s it produced
op
just a handful of movies each year. After fending off two hostile takeover bids, Disney named
Michael Eisner, a former television and movie executive, its CEO in 1984. Eisner and the head
of Disney’s movie studio, Jeffrey Katzenberg, increased both the quantity and quality of Disney’s
movies. Disney and its movie-studio subsidiaries—Touchstone Pictures (founded in 1984) and
Hollywood Pictures (founded in 1989)—produced roughly 10 movies per year by the late 1980s,
and between 20 and 30 annually through most of the 1990s.2 Touchstone and Hollywood produced
tC

live-action films intended for more mature audiences, with hits that included Pretty Woman and
The Sixth Sense. Eisner and Katzenberg also ushered in a second golden age of Disney animated
movies that extended from the late 1980s through the mid-1990s. The string of blockbusters in
this period included The Little Mermaid (1989), Beauty and the Beast (1991), Aladdin (1992),
The Lion King (1994), and Pocahontas (1995). Building on these successes, Disney launched two
book publishing houses—Disney Press (1990), a publisher of children’s books, and Hyperion
No

Books (1991), which published general-interest fiction and non-fiction books—and established a
thriving presence on Broadway. The musical version of Beauty and the Beast opened in 1994 and
remained in production until 2007. The Lion King, which opened in 1997, won six Tony awards
and went on to become the third-longest-running musical on Broadway.3

The era of aggressive expansion in Eisner’s first decade culminated in 1995 with Disney’s
acquisition of the media company Capital Cities/ABC. That deal included not only the TV
network ABC but also the cable sports network ESPN and multiple newspapers, magazines, and
Do

radio stations. Aside from a few movies that Disney produced in partnership with Pixar, however,
its string of film successes faded by the late 1990s. With the value of Disney’s shares stagnating,
Eisner was pushed out in 2005.

2 K e ll o g g S c h o o l of Management

This document is authorized for educator review use only by ARPITA SRIVASTAVA, GL Bajaj Institute of Management & Research (GLBIMR) until Aug 2023. Copying or posting is an
infringement of copyright. [email protected] or 617.783.7860
 KE1251 Disney+ and Machine Learning

t
Bob Iger, the president and COO of Capital Cities/ABC, succeeded Eisner as Disney’s CEO.

os
Iger prioritized increasing the company’s branded content, with a special focus on reinvigorating
the animation division. He wooed Pixar’s CEO (and Apple’s co-founder), Steve Jobs, who agreed
to sell Pixar to Disney for $7.4 billion in 2006. That was the first of four blockbuster deals during
Iger’s tenure. In 2009, Disney bought the Marvel comics franchise for $4 billion, giving the
company the rights to intellectual property such as the characters of Iron Man, Thor, Incredible

rP
Hulk, the Avengers, and Captain America.4 In 2012, Disney bought Lucasfilm for another $4
billion, giving it the rights to the Star Wars movie franchise.5 With the fourth and final blockbuster
deal of Iger’s tenure—the acquisition of 21st Century Fox in 2019—Disney added the movie
studios 20th Century Fox and Fox Searchlight Pictures, the cable network FX, and a controlling
stake in the on-demand streaming service Hulu. That $71.3 billion acquisition also gave it a vast
catalog of content—including movies dating to the 1930s and the rights to Marvel’s Fantastic

yo
Four and X-Men franchises, which Fox still held the rights to.

In 2019—just before the COVID-19 pandemic temporarily shut down much of the movie
industry—Disney produced seven of the top-ten-grossing movies in US theaters. Four were directly
tied to these blockbuster deals: Avengers: Endgame (Marvel), Toy Story 4 (Pixar), Captain Marvel
(Marvel), and Star Wars: The Rise of Skywalker (Lucasfilm). The other three were sequels/remakes
involving blockbusters by Disney Animation Studios: The Lion King, Frozen 2, and Aladdin.
op
As it had done since the 1950s, Disney cultivated and leveraged synergies between its film
and theme park divisions. In 2016, it opened Shanghai Disneyland, its sixth major theme park,
which followed Disneyland, Walt Disney World, Tokyo Disney Resort (1983), Disneyland Paris
(1992), and Hong Kong Disneyland Resort (2005). Disney also initiated a $24 billion program to
upgrade and expand its parks globally from 2018 to 2023. Many of these plans built on Disney’s
tC

IP acquisitions during the Iger era, including Marvel superhero lands at two parks, an Iron Man
rollercoaster at Disneyland Paris, and an expansion of Toy Story Land at Hollywood Studios, near
Walt Disney World.6 In 2018, Disney’s parks and resorts division generated nearly $20 billion
in revenue and about $4.5 billion in operating income. (See Exhibit 2 for Disney’s revenues and
operating profits by division.)
Iger’s last major initiative at Disney involved leading it into the market for video-on-demand
No

streaming. In 2012, Disney signed a deal that gave Netflix the right to stream its theatrically
released films. The deal took effect in 2016. In 2017, however, Disney announced that it would
launch two new streaming services, ESPN+ and Disney+, and pull its content from Netflix. That
decision coincided with a broader shift toward streaming among US TV viewers. From 2017 to
2019, for example, the percentage of people who subscribed to more than one streaming service in
the US nearly doubled, from 24 percent to 45 percent.7

Disney acquired a controlling interest in the streaming platform developed by Major League
Baseball, BAMTech, as a vehicle to launch its new streaming service. In 2018, it introduced
Do

ESPN+. Building on that launch—and on the IP acquired in its recent deals—Disney planned
to roll out Disney+ in the fall of 2019 with a content archive of roughly 7,500 TV episodes
and 500 films.8 It also planned to make major investments in original programming—starting
with a Star Wars spinoff, The Mandalorian, scheduled for release in December 2019. Disney had

K e ll o g g S c h o o l of Management 3

This document is authorized for educator review use only by ARPITA SRIVASTAVA, GL Bajaj Institute of Management & Research (GLBIMR) until Aug 2023. Copying or posting is an
infringement of copyright. [email protected] or 617.783.7860
 Disney+ and Machine Learning KE1251

t
the potential advantage of an established, video-on-demand subscriber base of about 27 million

os
people via Hulu, the streaming service it had gained in the Fox deal, which created opportunities
for bundling Disney+ and Hulu at a discounted price.9

As Gupta thought about this history and what it might meant for Disney’s path forward, her
thinking was guided by some basic facts and norms regarding ML’s capabilities, implementation,

rP
and uses.

Machine Learning and Data

“Artificial intelligence” (AI) was the umbrella nomenclature for multiple computer-enabled
activities, including robotics, image recognition, speech comprehension, and machine learning

yo
(ML). Although the British mathematician Alan Turing pioneered the concept of AI in the 1950s,
the computing power, algorithm development, and data availability to make it practical and
broadly available did not emerge until the 2010s.

ML dealt specifically with using historical data to make predictions. By analyzing vast amounts
of data and input variables, and leveraging sophisticated algorithms to find patterns, ML could be
used to develop models that create value for businesses. At the highest level, it answered three
op
questions:

1. Which of these things are similar? (i.e., descriptive, classification, segmentation, or clustering
analytics). A business use case could be knowing the shared characteristics of the most
valuable customers.
2. Which are most likely to act in a certain way? (i.e., predictive analytics). A business use case
tC

could be knowing which customers are most likely to buy in the near term.
3. How can we affect outcomes? (i.e., prescriptive or causal analytics). A business use case could be
knowing what to do to increase purchases.

Scientists answered these questions using a combination of data and algorithms to construct
models. The algorithms ranged from simple regressions to deep-neural networks that mimicked
how the human brain processed information. They also varied by the opaqueness of the decision-
No

making process. The program used by Gupta’s team could run multiple algorithms simultaneously
on the data, then provide statistics on the fit and accuracy of each algorithm for the given task.

Data about the attributes associated with an entity was critically important to the algorithm’s
functioning. If that entity was a machine on the shop floor, for example, the relevant data might
include its age, manufacturer, temperature, run time, throughput, and vibration. Any entity was
“tagged” with attributes. Some tags were manually created by humans, some were automatically
collected, and some were assigned via other AI. Also available was metadata—an underlying layer
Do

of information that described the data, often in a form that made it easier to use.

Customer data was extremely valuable—typically, the more, the better. Companies gathered
data about their current and potential customers so they could leverage ML tools and create new
value. Data could come from any of the following three types of sources:

4 K e ll o g g S c h o o l of Management

This document is authorized for educator review use only by ARPITA SRIVASTAVA, GL Bajaj Institute of Management & Research (GLBIMR) until Aug 2023. Copying or posting is an
infringement of copyright. [email protected] or 617.783.7860
 KE1251 Disney+ and Machine Learning

t
• First-party data about its users was collected directly by a firm—for example, through a

os
website, survey, or loyalty program.
• Second-party data about its users was collected directly by a firm, then made available to
other firms. The extensive user data that Google, Facebook, and Amazon collected—then
made available to advertisers—was a prime example.

rP
• Third-party data was aggregated by firms that did not originally collect it. Examples
included firms that compiled lists of people whose auto lease or cell phone plan would
soon expire, then sold that information to interested companies.

When Gupta researched examples to draw on and incorporate into her presentation to Disney
executives, Netflix stood out. Few companies rivaled its use of technology to build its customer
base over the past decade. She did some research into the specifics of its success.

yo
Netflix and ML
American entrepreneur Reed Hastings co-founded Netflix in 1997 as a mail-based DVD rental
company, driven in part by his frustration with paying late fees for video rentals at brick-and-
mortar businesses. Under the initial model, users could select new DVDs once they mailed back
op
the ones they had already rented, with no time limits or fines imposed. Hastings’s background as
a mathematician and engineer shaped the company from the outset, leading it to prioritize the
development and use of technology that could improve users’ experience and encourage them to
use Netflix regularly.10

In 2000, Netflix introduced a recommendation system that incorporated data from users’
tC

profiles (i.e., information they supplied about the categories of movies they liked), their movie
ratings, and the ratings of similar users. In 2006, it announced a $1 million prize for anyone who
could develop an algorithm that improved the accuracy of Netflix’s recommendation system by
10 percent. It awarded the prize in 2009—but never incorporated the algorithm. By that point,
Netflix was primarily focused on building out the video-on-demand streaming platform it had
introduced in 2007 as an alternative to its original, mail-based system. The value added by the
winning algorithm “did not seem to justify the engineering effort” required to incorporate it, as
No

two Netflix engineering and “personalization” scientists explained. “Also, our focus on improving
Netflix personalization had shifted to the next level by then.”11

Through the 2010s, Netflix played an increasingly active role in producing the films and shows
in its content catalog. In 2013, it became a distributor of first-run, original TV shows by acquiring
the rights to House of Cards and Orange Is the New Black. It expanded into studio-like production
with its first original TV series in 2015; the first Netflix-produced TV series that became a hit,
Stranger Things, debuted in 2016. From 2013 to 2017, Netflix’s planned spending on original
Do

content increased from $2.4 billion to $6 billion.12 By 2019, this figure had risen to an estimated
$15 billion.13

As it expanded, Netflix continued to rely heavily on ML to improve the accuracy of its

recommendations, the quality of its content, and users’ overall experience, investing nearly 10

K e ll o g g S c h o o l of Management 5

This document is authorized for educator review use only by ARPITA SRIVASTAVA, GL Bajaj Institute of Management & Research (GLBIMR) until Aug 2023. Copying or posting is an
infringement of copyright. [email protected] or 617.783.7860
 Disney+ and Machine Learning KE1251

t
percent of its annual revenue in technology and research in 2015. One analyst, who noted that

os
Netflix was hiring about three times more engineers than HBO was, predicted that the company
was “bound to figure out programming before its competitors—media companies—can fully
grasp the digital world, streaming technology, and the cloud.”14 In 2018, Fast Company ranked
Netflix the second most innovative company in the world, just behind Apple.15 Examples of its
innovations included a 2017 coding upgrade that allowed some scenes to stream more efficiently

rP
on phones, which helped minimize delays and improve the overall viewing experience. In 2018,
Netflix experimented with interactive, “choose-your-own-adventure” programming with Black
Mirror: Bandersnatch, a sci-fi movie with multiple endings. Bandersnatch won the 2019 Emmy
Award for Outstanding Television Movie.

Netflix relied heavily on ML to drive its growth and retain subscribers—with great success.
(See Exhibit 3 for Netflix’s growth trajectory from 2007 to 2016.) Most notably, it used ML

yo
to generate content recommendations, which accounted for an estimated 80 percent of the TV
shows that people watched on the service.16 As Netflix became primarily a streaming service—
and moved away from its original, mail-based model—its recommendation system increasingly
relied on inferences from users’ behavior. The data it collected included not only the content users
watched but how and when they watched it, what they watched immediately before and after, and
a wide variety of other behaviors.
op
Netflix also assigned staff to watch shows and tag them, scene by scene, with a range of
attributes, including details about the plot, the setting, and the cast. Then it combined these
tags with the behavioral data and used ML to determine the importance of each data point in
generating recommendations. As the Netflix VP of product innovation explained in a 2017 Wired
article: “How much should it matter if a consumer watched something yesterday? Should that
tC

count twice as much or ten times as much compared [with] what they watched a whole year ago?
How about a month ago? How about if they watched ten minutes of content and abandoned it or
they binged through it in two nights? How do we weight all that? That’s where machine learning
comes in. What those things create for us is ‘taste communities’ around the world. It’s about
people who watch the same kind of things that you watch.”17

A typical viewer would fit into several of the roughly 2,000 taste communities Netflix identified.
No

One great benefit of this approach—that is, inferring tastes from the behavior of users and of
people with similar tastes, rather than relying on the information users intentionally supplied—
was that the system could identify shows, movies, and categories of content that were not obvious
as good fits. In 2015, for example, one in eight people who watched a Marvel show on Netflix had
never before watched a comic book-based show on the service.18 In most cases, they did so likely
because of a Netflix recommendation.

Netflix also used ML to understand “the characteristics that make content successful,” as a
company blog post put it.19 In practice, this meant that Netflix’s creative directors and executives
Do

sometimes gave feedback regarding the development of scripts based on data rather than simply
intuition. The writer and director Cary Fukunaga described in a GQ magazine article how this
process played out in Maniac, a 10-episode miniseries that debuted on Netflix in 2018. “Because
Netflix is a data company, they know exactly how their viewers watch things,” Fukunaga said. “So

6 K e ll o g g S c h o o l of Management

This document is authorized for educator review use only by ARPITA SRIVASTAVA, GL Bajaj Institute of Management & Research (GLBIMR) until Aug 2023. Copying or posting is an
infringement of copyright. [email protected] or 617.783.7860
 KE1251 Disney+ and Machine Learning

t
they can look at something you’re writing and say, ‘We know based on our data that if you do this,

os
we will lose this many viewers.’ So it’s a different kind of note-giving. It’s not like, Let’s discuss this
and maybe I’m gonna win. The algorithm’s argument is gonna win at the end of the day. So the
question is do we want to make a creative decision at the risk of losing people.”20

Netflix’s ML had the potential to subordinate creative risks to the goal of reaching and retaining

rP
a larger audience, as Fukunaga suggested. But in some ways, it could also promote risk-taking and
creativity. As the Harvard Business Review noted, “digital platforms like Netflix have more data on
consumer tastes than any entertainment company has before,”21 but the data’s value had little to do
with the fact that it gave Netflix leverage to micromanage content—which it rarely did. Instead,
Netflix was “in the business of assembling the best collection of movies and shows to meet the
needs of each viewer.”22 Consequently, a movie or show’s potential to be a blockbuster was less
important than its power to connect with relatively small taste communities. That kind of capacity

yo
for precision matchmaking could, in general, create great value for both consumers and businesses,
according to the authors.23

Industry analysts pointed to this model as a key competitive advantage because it created
efficiencies in the show development and production process. By connecting viewers with shows
they would like, Netflix achieved a renewal rate for original shows that was roughly three times the
rate of legacy TV networks. These efficiencies—and a high subscriber retention rate—saved the
op
company more than $1 billion annually, by Netflix’s estimate.24 The evidence also suggested that
the model gave writers and directors the leeway they needed to produce high-quality shows: Netflix
was nominated for 112 Emmy Awards in 2018, the most of any network or streaming service.
Second-place HBO, with 108 nominations, had topped the list for 17 consecutive years before
Netflix’s achievement. It was “a tangible awards victory for Netflix’s flood-the-market strategy,” The
tC

Atlantic noted, “and it’s barely getting started.”25

Netflix’s primary use of ML, beyond content recommendations and development, involved
individualizing and improving the user experience. For example, the Netflix home page thumbnail
images for selecting shows were personalized per viewer, based on Netflix’s data about a user’s taste
preferences and about the characteristics of effective thumbnails. The company’s research showed
that users typically spent no more than 90 seconds browsing the site before moving on if they
No

found nothing that interested them. Research also showed that users spent 82 percent of their time
focusing on artwork; that they spent a little under two seconds on each image; and that images
were the single biggest factor in what they chose to watch. Beginning in 2014, Netflix spent two
years building a system that created multiple thumbnails for each show or movie, then displayed
the image that was most likely to motivate that particular viewer to try the show. “In the end,” a
Netflix analyst wrote, “we saw one clear thing: Using better images to represent content significantly
increased overall streaming hours and engagement from our members.”26 The common qualities of
effective thumbnail images, according to Netflix’s research, included expressive faces and complex
Do

emotions, villainous characters, and no more than three people in an image.

ML also helped Netflix provide users with an improved viewing experience by identifying
and predicting periods of high demand and employing adaptive technologies that minimized lags

K e ll o g g S c h o o l of Management 7

This document is authorized for educator review use only by ARPITA SRIVASTAVA, GL Bajaj Institute of Management & Research (GLBIMR) until Aug 2023. Copying or posting is an
infringement of copyright. [email protected] or 617.783.7860
 Disney+ and Machine Learning KE1251

t
in the streaming experience.27 In 2019, Netflix accounted for nearly 13 percent of downstream

os
streaming traffic on the internet, about twice that of Google’s YouTube.28

Decision Point
Heading to bed, Gupta thought about how Netflix again had kept her glued to the TV later

rP
than she intended or wanted. And she thought about her meeting in three days. How could she
explain the potential value of ML to senior executives? Some applications for Disney+ could be
simply ported from Netflix, but her audience likely already knew about those. What value could
Gupta and her team add—and what kind of data would they need to leverage? And what, if any,
ethical concerns might arise from gathering such data? Because algorithms do only what humans
train them to do, was there an inherent bias in the data that would discriminate against certain

yo
customers—as when Amazon found that its experimental AI-driven hiring tool discriminated
against women?29 That kind of bad press was the last thing the new launch needed.

As she thought about these unknowns and challenges, Gupta was clear-eyed and confident
about one thing: Disney was about to engage with a field of stellar competitors—including some
tech and entertainment companies it had not directly competed with until now. Without doubt,
success in this new era would depend on making the most of ML.
op
tC
No
Do

8 K e ll o g g S c h o o l of Management

This document is authorized for educator review use only by ARPITA SRIVASTAVA, GL Bajaj Institute of Management & Research (GLBIMR) until Aug 2023. Copying or posting is an
infringement of copyright. [email protected] or 617.783.7860
 KE1251 Disney+ and Machine Learning

t
Exhibit 1: 1957 Map of Disney’s Activities

os
rP
yo
op
tC

Source: “Disney’s 60-Year-Old Growth Map Answers the Netflix Question,” Reforge, accessed August 24, 2022, https://www.reforge.
com/blog/disney-growth-map.
No
Do

K e ll o g g S c h o o l of Management 9

This document is authorized for educator review use only by ARPITA SRIVASTAVA, GL Bajaj Institute of Management & Research (GLBIMR) until Aug 2023. Copying or posting is an
infringement of copyright. [email protected] or 617.783.7860
 Disney+ and Machine Learning KE1251

t
Exhibit 2: Disney’s Revenue and Operating Income by Segments, 2012–

os
2018

rP
yo
op
tC
No
Do

Source: Oleksandr Pylypenko, “The Walt Disney Company: What’s in the Cards in 2019?,” Seeking Alpha, February 1, 2019, https://
seekingalpha.com/article/4237424-walt-disney-company-in-cards-in-2019.

10 K e ll o g g S c h o o l of Management

This document is authorized for educator review use only by ARPITA SRIVASTAVA, GL Bajaj Institute of Management & Research (GLBIMR) until Aug 2023. Copying or posting is an
infringement of copyright. [email protected] or 617.783.7860
 KE1251 Disney+ and Machine Learning

t
Exhibit 3: Increase in Netflix Subscribers, 2007–2016

os
rP
yo
op
Source: Jeff Dunn, “Here’s How Huge Netflix Has Gotten in the Past Decade,” Insider, January 19, 2017, https://www.businessinsider.
tC

com/netflix-subscribers-chart-2017-1.
No
Do

K e ll o g g S c h o o l of Management 11

This document is authorized for educator review use only by ARPITA SRIVASTAVA, GL Bajaj Institute of Management & Research (GLBIMR) until Aug 2023. Copying or posting is an
infringement of copyright. [email protected] or 617.783.7860
 Disney+ and Machine Learning KE1251

t
Endnotes

os
1 Talib Visram, “Netflix Spent an Estimated $15 Billion on Original Content in 2019, Ahead of Disney+ and Apple
TV+’s Launches,” Fast Company, November 8, 2019, https://www.fastcompany.com/90410798/netflix-spent-an-
estimated-15-billion-on-original-content-in-2019-ahead-of-disney-and-apple-tvs-launches.
2 “List of Disney Films,” D23, accessed August 21, 2022, https://d23.com/list-of-disney-films.

rP
3 Sally Henry, “The Lion King Becomes 3rd-Longest-Running Broadway Show of All Time; Passes Cats,” Broadway
World, October 31, 2015, https://www.broadwayworld.com/article/THE-LION-KING-Becomes-3rd-Longest-
Running-Broadway-Show-of-All-Time-Passes-CATS-20151031.
4 “Disney to Acquire Marvel Entertainment,” press release, August 31, 2009, https://thewaltdisneycompany.com/
disney-to-acquire-marvel-entertainment.
5 “Disney to Acquire Lucasfilm Ltd.,” press release, October 30, 2012, https://thewaltdisneycompany.com/disney-
to-acquire-lucasfilm-ltd.
6 Brooks Barnes, “Disney Is Spending More on Theme Parks than It Did on Pixar, Marvel and Lucasfilm Combined,”

yo
New York Times, November 16, 2018, https://www.nytimes.com/interactive/2018/11/16/business/media/disney-
invests-billions-in-theme-parks.html.
7 “Video Streaming in the US: An Industry in Flux,” Statista, 2019, 10.
8 Ibid., 15.
9 Visram, “Netflix Spent an Estimated $15 Billion.”
10 Trey Williams, “Netflix CEO Wants to Take Over the World,” MarketWatch, July 31, 2015,
https://www.marketwatch.com/story/can-netflix-achieve-its-goal-of-world-domination-2015-07-30.
op
11 Xavier Amatriain and Justin Basilico, “Netflix Recommendations: Beyond the 5 Stars (Part 1),” Netflix
Technology Blog, April 6, 2012, https://netflixtechblog.com/netflix-recommendations-beyond-the-5-stars-part-1-
55838468f429.
12 Daniel Sparks, “Netflix Has an Enormous Content Budget,” Insider, September 4, 2017,
https://www.businessinsider.com/netflix-has-an-enormous-content-budget-2017-9.
13 Visram, “Netflix Spent an Estimated $15 Billion.”
tC

14 Williams, “Netflix CEO Wants to Take Over the World.”

15 “The World’s 50 Most Innovative Companies 2018,” Fast Company, accessed January 23, 2023,
https://www.fastcompany.com/most-innovative-companies/2018.
16 Libby Plummer, “This Is How Netflix’s Top-Secret Recommendation System Works,” Wired, August 22, 2017,
https://www.wired.co.uk/article/how-do-netflixs-algorithms-work-machine-learning-helps-to-predict-what-
viewers-will-like.
17 Ibid.
No

18 Ibid.
19 “Research Areas: Machine Learning,” Netflix Research, accessed August 19, 2022, https://research.netflix.com/
research-area/machine-learning.
20 Zach Baron, “Cary Fukunaga Doesn’t Mind Taking Notes from Netflix’s Algorithm,” GQ, August 27, 2018,
https://www.gq.com/story/cary-fukunaga-netflix-maniac.
21 Michael D. Smith and Rahul Telang, “Data Can Enhance Creative Projects—Just Look at Netflix,” Harvard
Business Review, January 23, 2018, https://hbr.org/2018/01/data-can-enhance-creative-projects-just-look-at-
netflix.
Do

22 Ibid.
23 Ibid.
24 Trey Williams, “Netflix Uses Frame-by-Frame Machine Learning to Decide What You Really Want to Watch,”
MarketWatch, September 28, 2017, https://www.marketwatch.com/story/netflix-uses-frame-by-frame-machine-
learning-to-decide-what-you-really-want-to-watch-2017-09-27.

12 K e ll o g g S c h o o l of Management

This document is authorized for educator review use only by ARPITA SRIVASTAVA, GL Bajaj Institute of Management & Research (GLBIMR) until Aug 2023. Copying or posting is an
infringement of copyright. [email protected] or 617.783.7860
 KE1251 Disney+ and Machine Learning

t
25 Sophie Gilbert, “Emmy Nominations 2018: Netflix Takes Over,” Atlantic, July 12, 2018,

os
https://www.theatlantic.com/entertainment/archive/2018/07/emmy-nominations-2018-netflix-hbo/565046.
26 Nick Nelson, “The Power of a Picture,” Netflix, May 3, 2016, https://about.netflix.com/en/news/the-power-of-a-
picture.
27 “Research Areas: Machine Learning,” Netflix Research.
28 Todd Spangler, “Netflix Bandwith Consumption Eclipsed by Web Media Streaming Applications,” Variety,

rP
September 10, 2019, https://variety.com/2019/digital/news/netflix-loses-title-top-downstream-bandwidth-
application-1203330313.
29 Jeffrey Dustin, “Amazon Scraps Secret AI Recruiting Tool That Showed Bias Against Women,” Reuters, October
10, 2018, https://www.reuters.com/article/us-amazon-com-jobs-automation-insight/amazon-scraps-secret-ai-
recruiting-tool-that-showed-bias-against-women-idUSKCN1MK08G.

yo
op
tC
No
Do

K e ll o g g S c h o o l of Management 13

This document is authorized for educator review use only by ARPITA SRIVASTAVA, GL Bajaj Institute of Management & Research (GLBIMR) until Aug 2023. Copying or posting is an
infringement of copyright. [email protected] or 617.783.7860