Module 3. Netflix - Leveraging Big Data

ITSY/075
IBS Center for Management Research
Netflix: Leveraging Big Data to Predict Entertainment Hits

This case was written by Tangirala Vijay Kumar, under the direction of Debapratim Purkayastha,
IBS Hyderabad. It was compiled from published sources, and is intended to be used as a basis for
class discussion rather than to illustrate either effective or ineffective handling of a management
situation.
 2013, IBS Center for Management Research. All rights reserved.
To order copies, call +91 9640901313 or write to IBS Center for Management Research (ICMR), IFHE Campus, Donthanapally,
Sankarapally Road, Hyderabad 501 504, Andhra Pradesh, India or email: [email protected]
www.icmrindia.org
License to Use for IBS Campuses Only. Sem II. Class of 2017-2019.
ITSY/075

“Netflix has so thoroughly analyzed viewer patterns that they now know what viewers enjoy and
don’t enjoy — before viewers even see the show.”1
– Joe Kukura, in February 2013.
“We’ve entered an era when product decisions can be made not by analysis of demographics or
user testing but by extremely fine-tuned measurements of current user activities designed to
predict what new activities they will be eager to engage in.”2
– Damian Rollison, Vice President of Product and Technology, Universal Business
Listinga, in February 2013.
“My worry about BD (big data) was that it will steal spontaneity from our lives by figuring out
what we like before we’ve even had a chance to decide.”3
– Mike Cassidy, in February 2013.
Netflix, a company which operated in the DVD-by-rental segment and in the video streaming
segment, had, since its inception, garnered significant insights into the viewing patterns of its
customers. It knew what shows they preferred viewing, the time at which they watched the shows,
the devices on which they streamed the videos, and other such viewing nuances. According to
observers, the company’s big data capabilities had received a shot in the arm from the cloud
computing infrastructure it sourced from Amazon Web Services. Netflix had initially employed
this data to recommend movies and TV shows to its subscribers. However, with increased
competition for broadcast rights of movies and TV shows which had their initial run and the
reluctance on the part of some potential competitors to equip Netflix with movie and TV program
ammunition, Netflix plunged into original programming. The company clinched the exclusive
rights to the first and second seasons of the show “House of Cards” for a staggering US$100
million. According to industry observers, the company bid this amount after confirming that many
of its subscribers were fans of the director, David Fincher, and the lead actor of the show, Kevin
Spacey, and had been followers of the original “House of Cards” showb. The show proved to be a
roaring success, but the most incredible aspect, according to observers, was that Netflix executives
were sure that it would succeed even before the first scene was shot.
However, some industry observers were concerned that Netflix would find itself on the wrong side
of the law with privacy concerns being raised by some observers over the employment of big data
capabilities to gauge the personal viewing habits of its subscribers. Others were apprehensive
about the potential damage to Netflix’s reputation due to its site being down on account of outages
in Amazon Web Services. Analysts were also concerned about Netflix stretching itself financially
to create a credible image for itself through original programming. Some industry observers were
apprehensive that as companies such as Netflix, which had greater awareness about the audience’s
viewing habits, gained more prominence in the generation of original programming, they might
begin influencing the creative decisions of even directors and writers.4
a
Universal Business Listing is a company focused on promoting online visibility for local businesses.
b
House of Cards was based on a 1989 novel by British writer Michael Dobbs. It revolved around one
politician’s ruthless methods to clinch power and handicap his rivals. The original “House of Cards” was
a mini-series which was broadcast by the BBC in 1990.
1
BACKGROUND NOTE
Netflix was started in 1997. In 1999, the company introduced a DVD subscription service, providing
unlimited rentals on a monthly subscription basis. In 2000, it introduced the personalized movie
recommendation mechanism that employed subscribers’ ratings to forecast the preferences of all
Netflix subscribers.5 Toward the close of 2007, Netflix was shipping around 1.6 million DVDs on a
daily basis.6 In 2007, the company also launched its video streaming services in the US, and, in 2010,
began the global expansion of this business. As of March 2013, Netflix, Inc. functioned in three
categories: Domestic Streaming, International Streaming, and Domestic DVD.
The Domestic DVD business provided DVDs-by-mail subscription services in the US. Under the
streaming services, Netflix provided an Internet television network service that allowed
subscribers to stream TV programs and films directly on TVs, computers, and mobile gadgets in
the US and globally. 7 As of December 31, 2012, Netflix had 27.15 million subscribers for its
streaming service in USA. Subscribers in other parts of the world totaled 6.12 million. The
subscribers to Netflix’s mail order business totaled 8.22 million.8 As of November 2012, Netflix
accounted for 33% of the peak period video streaming traffic in North America.9 As of March
2013, the company’s subscribers in nearly 40 nations were viewing close to 1 billion hours of TV
shows and movies on a monthly basis.10 As of March 2013, Netflix was charging US$ 7.99 per
month from each of its video streaming subscribers in USA. For the year ended December 31,
2012, Netflix had generated revenues of US$3,609.3 million and net income of US$17.1 million
(Refer to Exhibit I for Netflix’s financial performance from 2005 to 2012).
BIG DATA AT NETFLIX

Netflix heralded the “big data” approach to Internet media, wherein each likely metric of experience
was recorded, assessed, and inventoried forever (Refer to Exhibit II for a brief note on big data).11
TRACKING SUBSCRIBERS’ VIEWING PATTERNS
Netflix tracked each search that a viewer made, each good or bad rating attributed by a viewer to
what s/he had just seen, besides the ratings data from third-party providers such as Nielsen. These
were in addition to the location data, device data, remarks on the social media, etc. Netflix was
also aware of what a subscriber most probably viewed on each of his/her devices like mobile or
laptop or tablet and what viewers belonging to a specific ZIP code preferred viewing on their
laptops on a Saturday night. It even kept a tab on the number of subscribers who switched off
before the credits began to roll. 12 The company also captured how a subscriber watched programs,
including the number of times s/he paused the show, whether s/he was a binge viewer, whether
s/he often exited the show to view others, and, whether s/he left shows incomplete. 13 Explaining
how Netflix’s ability to gauge viewer preferences had evolved over time, Ted Sarandos
(Sarandos), Netflix’s Chief Content Officer, stated, “Here is what the data from our DVD business
tells us: we know what we shipped to you and we know when you returned it. I have no idea if you
watched it. I have no idea if you watched it 20 times. With streaming, we have insight into every
second of the viewing experience. I know what you have tried and what you have turned off. I
know at what point you turned it off.”14
In gauging the preferences of its subscribers, Netflix also took into account their ratings, the
queuing up of movies/ shows by subscribers, and other viewer-driven actions. It combined these
findings with the type of content viewed by them, as for instance, action or comedy. 15 While the
significance of isolated incidents such as taking a bathroom break, sleep hours, etc., could be
minimized, Netflix was focused on the trends that began to become visible over time. According to
an information technology expert, this level of garnering data bestowed upon Netflix a level of
insight into its subscribers’ needs and preferences that none of the television or studio networks
could ever tap into (but which they all wanted, as was evident from the innumerable surveys and
viewer polls that Hollywoodc had administered over the years).16
c
The term Hollywood is used to refer to the American film industry and the lifestyles of the individuals
connected to it.
2
Netflix’s data about its subscribers had increased at least 10 times from 2007 till 2013, as it
focused more on the Internet streaming business than on its DVD rental business.17 As noted by an
information technology expert, Netflix’s strategy ultimately boiled down to “capture data about
‘What’ you offer your users, and ‘How’ they interact with it, then you can infer ‘Why’ things work
or don’t work.”18
INVENTORY OF DATA AND ALGORITHM
In the last quarter of 2011, Netflix’s streaming service subscribers had seen over 2 billion hours of
streaming video. As of June 2012, Netflix had over 25 million subscribers to its streaming service.
Given this subscriber base and usage, Netflix captured, among others, around 30 million user
actions on a daily basis such as rewinds, fast forwards, and pauses, 4 million ratings on a daily
basis, and, searches numbering 3 million on a daily basis.19
As of July 2011, Netflix was storing 50 distinct files for each viewable item, including 3 copies of
each film, viewer ratings for ten years, comprehensive viewer account information, and metadata
comprising intricate licensing rights for everything, audio files, log files, subtitles, etc. Netflix had
globally placed cloud data centers for minimum latencyd, verification data for subscribers, data
about the test celle that every viewer was in, and the gadget that each account activated,
bookmarksf, personalization, and category preferences. 20
Netflix compiled all of its subscribers’ viewing patterns and took all of its customers’ viewing
habits and show ratings and applied a sophisticated algorithm to them. 21 A Netflix blog post dated
April 6, 2012, stated, “Streaming has not only changed the way our members interact with the
service, but also the type of data available to use in our algorithms.”22 The algorithm employed by
Netflix derived data from multiple sources and factored in different behaviors. The star rating
given by viewers was an important source. If a subscriber rated a show without streaming it as s/he
had viewed it in a theater or on DVD or s/he rated it casually, the algorithm allotted a lower weight
to the rating. In case a subscriber saw only a part of the show and rated it, the weight given to the
rating got proportionally reduced.23
PERSONALIZED RECOMMENDATIONS
Netflix gave viewing suggestions by employing algorithms that came up with apt content before
the appropriate viewer with the help of data such as their prior viewing patterns and also taking
into account the viewing behaviors and recommendations of other subscribers.24 As of April 2012,
three-fourth of the shows viewed by subscribers resulted from some kind of recommendation. 25
The company was reputed for its recommendation and targeting technologies that helped it dish
out personalized content to the audience that had the highest probability of viewing it. If a
subscriber of Netflix’s streaming service wanted to watch a movie, the company employed its
mammoth product-attribute datasets and the ratings of viewers numbering millions to narrow down
on the appropriate movie. By assessing data regarding the movies that subscribers ordered for, it
could suggest the films that they could watch subsequently. Subscribers liked it as they got
recommendations about movies they would potentially be interested in. The benefit for Netflix was
that subscribers ordered for more movies.26 Explained Sarandos, “We’ve made really incredible
d
Latency implies any of various types of delays usually involved in processing of network data. (Source:
Bradley Mitchell, “Network Bandwidth and Latency,” http://compnetworking.about.com/od/speedtests/a/
network_latency.htm)
e
A test cell is a group of real and virtual machines with particular hardware and software configurations
that comply with the requirements of a particular test environment. Teams or individual testers can
reserve the cell to conduct tests. (Source: Pietro Marella, “Extend Rational Quality Manager to Manage
Lab Assets in the Cloud,” www.ibm.com/developerworks/rational/library/quality-manager-manage-lab-
assets-cloud/index.html)
f
Bookmark is the address of a web page that is saved on the computing device so that the user can access
it again easily.
3
strides to predict what people watch right after they finish something else. Basically, it’s a
statistical push based on what other people have watched and really enjoyed immediately after
viewing the same program you just finished.”27
The homepages of Netflix for each of its subscribers were personalized to the maximum extent
possible. The homepage comprised groups of videos arrayed in horizontal rows. Every row had a
label that indicated the intended meaningful link among the videos in that cluster. A major portion
of Netflix’s personalization boiled down to the manner the company chose the rows, the way it
arrived at the videos to be a part of the rows, and upon the sequencing of the videos. A common
basis for personalization was the compilation of rows based on genres like comedy or drama.
Every row comprised three tiers of personalization: the selection of the genre itself, the sub-
category of videos included in that genre, and the sequencing of those videos. The basis for
arriving at the genre preferences of a subscriber were the recent videos s/he had viewed, his/ her
ratings, and his/ her actions such as stopping the show midway, or direct feedback garnered
through viewing preferences survey. Similarity was also a key basis for personalization in Netflix
delivering its service. Similarity was perceived in different ways, for instance, in searching or
including videos in the queue. 28
The basic ranking model employed by Netflix for each individual balanced his/ her predicted
rating and the show/ movie’s popularity (Refer to Exhibit III for the basic ranking model employed
by the company). Netflix further fine-tuned its findings by employing A/B testingg (Refer to
Exhibit IV for visual representation of the testing which was carried out). 29 According to experts, a
key part of the strategy that Netflix employed to provide a rich experience to its viewers in terms
of user interactions and recommendations was the software architecture that it had put in place
(Refer to Exhibit V for Netflix’s software architecture).
Todd Yellin, Vice President of Product Innovation at Netflix, detailed the intricate process by
which Netflix arrived at the suggestions for its Watch Instantly menu. According to him, “One of
the challenges of big data is we’ll have all these terabytes flowing at us, and we have to figure out
which data to pick.”30 This, according to him, implied fragmenting the data further into, for
instance, what a specific viewer watched the previous night and what s/he watched the previous
year. He pointed out that what a subscriber viewed in the recent past was of more value. Then
Netflix combined it with other information such as ratings, if any, ascribed by the viewer at any
point in time and also the likely demographics of the family constituting the account. In the
summer of 2012, Netflix also introduced a “post play” feature wherein after one of the episodes of
a show ended and credits rolled, a pop-up window suggested to the subscribers that they view the
subsequent episode. The timing of the pop-up was carefully adjusted by an algorithm which
worked out when a viewer would most probably switch off. If a viewer did not engage in any
actions, the subsequent episode began automatically.31
SHIFT FROM OWN DATA CENTERS TO AMAZON WEB SERVICES

According to observers, Netflix’s big data capabilities received a boost with the gradual shift,
beginning in 2009, of the company’s computing infrastructure from its own data centers to the
cloudh, by engaging the services of Amazon Web Services (AWS), the cloud computing subsidiary
g
A/B testing is a simple technique of testing alterations to a webpage against the present design and
finding out which of the alterations generate positive results. It is a technique of verifying that any new
outlay or alteration to an element on a webpage is enhancing the conversion rate prior to making the
alteration to the site code. An A/B test comprises testing two versions of a web page — an A version (the
control) and a B version (the variation) — with real-time traffic and assessing the impact each version has
on the conversion rate. (Source: “What is A/B Testing?,” https://www.optimizely.com/ab-testing)
h
Cloud computing is an Internet technology that enables one to accomplish one’s work online without
having to set up hardware or software and without storing data at one’s end. A user can access any of his/
her applications through any device having an Internet browser and online connectivity.
4
of Amazoni. Netflix’s search and recommendation engine and its streaming servers started
utilizing AWS’s cloud infrastructure.32,33 The principal reason for Netflix’s shift was that its pace
of setting up data centers could not catch up with the increase in demand for its streaming
services.34 The convenience of scaling up information technology infrastructure on cloud, payment
for only the computing resources used, and allowing internal IT staff to concentrate on operations
of greater significance than just ensuring that servers were working fine were other reasons for
Netflix’s tapping the services of AWS.
Netflix employed Amazon Elastic MapReducej to assess streaming sessions and derive business
metrics regarding performance, viewing patterns, etc., which allowed it to continue to enhance the
quality of streaming.35 Hadoop’sk processing capabilities enabled Netflix to conduct huge data
analyses like creating visual representations of traffic patterns for each kind of device traversing
several markets. The increased processing competencies bestowed by Hadoop also enabled
Netflix’s engineers to find out where traffic on the network was slowing down, enabling it to order
for extra network capacity (Refer to Exhibit VI for Netflix’s cloud-based Hadoop architecture).
The technology—which could process bigger data sets — also aided Netflix to gauge customer
choices better so that it could come out with better recommendations.36 As of November 2012,
AWS was managing 95% of the computation and storage requirements of Netflix.37
THE NEED FOR ORIGINAL CONTENT AT NETFLIX

Prior to March 2011, Netflix had had no plans of becoming the first avenue where a TV show or a
movie was beamed. Rather, it concentrated on shows and movies having their second, third, or
even fourth runs. In other words, Netflix dealt with content subsequent to its release either in
theaters or on television. And occasionally, Netflix could lay its hands on content after it played in
theaters and subsequently on television for quite a while. 38 In July 2011, a prominent securities
company stated that Netflix required more streaming content to appeal to customers, and that only
around one-fifth of Netflix movies could be streamed online. At that time, a financial analyst had
observed, “The bigger Netflix gets, the more content deals they can do. But they’ll never be online
what they are in DVDs, because of rights and distribution windows…”39
The prime programming which was critical to Netflix’s streaming service was becoming difficult
to access.40 Content owners raised the fees they charged from Netflix and in some cases cancelled
their licensing arrangements with Netflix.41 HBO, a creator of well-known television shows,
declined to license any of its shows to Netflix. Comcast, a big cable company with a strong
portfolio of sports shows, also showed no inclination to join hands with Netflix. Showtime did not
license its shows before the passage of a couple of seasons. Industry observers expected such
tactics to leave Netflix with the option of only streaming old content. Recalled Sarandos, “I looked
at that and realized we were faced with a supply source that wasn’t reliable.”42 In the meantime,
Netflix was also facing increased rivalry. Hulul, HBO, Amazon.com, Comcast, Applem, Googlen,
i
Amazon.com, Inc. is an American multinational e-commerce company. It is an important player in the
business of providing cloud computing services as well.
j
Amazon Elastic MapReduce is a web service employing which a user can immediately provision as much
or as little capacity as he/ she requires to perform data-intensive operations for applications like web
indexing, data mining, log file analysis, data warehousing, machine learning, financial analysis, scientific
simulation, and bioinformatics research. It is not necessary to worry about issues such as time-consuming
setup and handling.. (Source: http://aws.amazon.com/elasticmapreduce/)
k
Hadoop or Apache Hadoop is an open-source software framework that helped the processing of large
data sets in a distributed computing environment. Amazon Elastic MapReduce made use of the Hadoop
framework.
l
Hulu is a website and over-the-top (OTT) subscription service providing ad-supported on-demand
streaming video of TV shows, films, webisodes and other new media, trailers, clips, and behind-the-
scenes footage from networks and studios. (Source: Daniel Ferry, “A Bear Gives a Bull Case for Netflix,”
www.fool.com, February 24, 2013)
5
and many start-ups were creating competing businesses. Analysts expected these companies to
clash with Netflix for content and also subscribers; they also expected them to vie with Netflix in
auctions for exclusive rights to television shows and movies. 43,44 Hence, Netflix’s CEO, Reed
Hastings, decided that the company would finance original shows on its own. Also, by creating
original content, Netflix expected to enjoy the type of credibility generally associated with an
AMCo or with HBO. Explained Sarandos, “If they like it they watch more. If they watch more,
they will value the service more.”45
ENTER THE ‘HOUSE OF CARDS’
IDENTIFYING THE SHOW

Netflix made viewers label films and TV shows with hundreds of metadatap descriptorsq to
comment, among others, about the actors, the screenplay, the overall appeal, and the category.
Previously, these labels were employed to suggest other shows available on the service,
fundamentally constructing outlines of individual viewers according to their choices. However,
Netflix began gradually ordering original content as it was aware of what subscribers desired prior
to their being in the know. 46
Netflix, akin to other networks, desired a series of hits. The challenge for it was never to have access
to viewers’ homes — its streaming service could be accessed on any gadget that could be connected
to television. The challenge was to create content that viewers were keen to watch and would not
have access to elsewhere. Rather than create lots of content and wishing that a part of it would be
lapped up by viewers, it utilized its huge data repository of 29 million subscribers’ viewer choices
and patterns.4748 When the show ‘House of Cards’, a remake of a BBC mini-series, was auctioned in
2011 with the combination of David Fincher and Kevin Spacey, Netflix executives mined their
mammoth data repository. The company’s assessment of viewer data helped it realize that a remake
of the famous BBC show on political intrigue could deliver the goods. Taking into account the
content and the team involved, Netflix was confident about the potential viewership being substantial
enough for the show to be a success.49,50 Unlike other networks which conventionally approved a
show subject to their being satisfied with a trial episode, Netflix approved two complete seasons (26
episodes) of House of Cards without viewing even one scene.51 It gained exclusive rights for the two
seasons reportedly for a price tag of over US$100 million.52
Netflix also employed data to target different viewers with different trailers. The total number of
distinct trailers numbered ten. The trailer beamed to a viewer was based on his/ her viewing history.
If a subscriber viewed action movies more, s/he was beamed a racy, energetic version. If a subscriber
viewed more of romances or dramas, s/he was streamed a trailer which emphasized characters.53
Netflix started making the program ‘House of Cards’ available on February 1, 2013, to the 33
million subscribers of its streaming service globally.54 ‘House of Cards’ was the first show wherein
the TV program rolled out the complete season comprising thirteen episodes at one time for
subscribers to view as per their convenience. 55
m
Apple Inc., is an American multinational company that designs, develops, and sells consumer electronics,
computer software, and personal computers. For the year ended September 29, 2012, Apple had
generated net sales of US$156,508 million and net income of US$41,733 million.
n
Google Inc. is an American multinational company specializing in Internet-related services and products.
These comprise search, cloud computing, software, and online advertising technologies. For the year
ended December 31, 2012, Google had generated total revenues of US$50,175 million and net income of
US$10,737 million.
o
AMC is an American cable and satellite television channel that is owned by AMC Networks. The channel
basically broadcasts theatrically released movies, coupled with a limited number of original shows.
p
Metadata represents a set of data that describes and provides information about other data.
q
A descriptor is an item of stored data that points out how other data is stored.
6
RESULTS
Netflix did not come out with data regarding either the number of its subscribers who viewed
House of Cards or viewer ratings of the show as it did not follow the business model of selling
advertising slots.56,57 However, based on circumstantial evidence, industry observers proclaimed
‘House of Cards’ to be a hit. Over 10,000 comments were posted on social media within one day
of ‘House of Cards’ debuting. According to social media analysts at the media research firm,
Fizziology, in the first 24 hours of its debut, House of Cards figured thousands of times per hour in
social media chats. Fizziology also observed that 62 percent of these comments were positive. 58
House of Cards garnered outstanding reviews with a total of 9 on IMDb.comr.59
Unmetric, a social media benchmarking company, stated that the trailer of House of Cards was the
most viewed video on Netflix’s YouTube Channel accounting for over 1.2 million views – one
million of these views were garnered from January 16 to February 8. According to experts, this
program was largely responsible for Netflix’s enhanced footprint on the social media. They also
opined that by attracting more viewers to Netflix’s YouTube Channel, House of Cards was
considerably expanding the company’s marketing avenues. Experts expected the show to aid
Netflix in increasing the number of its subscribers as it created additional content. According to
Trendrr, a company focused on deciphering social media data, 63% of the comments within the
first three days of debut of House of Cards i.e. from February 1 to February 3, was positive. 60 On
February 4, i.e. on the fourth day of the release of House of Cards, wired.com conducted a poll of
its readers to view how many among them had viewed House of Cards. Of the 1,900 readers who
participated in the poll, over 26% confirmed that they had watched all the 13 episodes.61
On February 12, 2013, Sarandos declared that House of Cards was the most-viewed show, both in
terms of number of distinct viewers and overall hours streamed, in each of the countries where
Netflix offered its streaming services. 62 He also expected the number of viewers of the show to
increase with the passage of time. According to him, Netflix would expand its portfolio of original
shows as it would give it bargaining power when negotiating with studios on getting exclusive
access to original shows. According to him, this did not, however, mean an increase in Netflix’s
subscription fees given the company’s subscriber volumes.63 According to an investment analyst,
House of Cards bestowed upon Netflix the credibility that it was capable of producing original
content on its own. The company would have a receptive audience for any of the original shows
that it came up with.64
THE REASONS BEHIND THE SHOW’S SUCCESS
According to observers, the success of House of Cards was not a freak incident. To collect real-
time data about the programs its subscribers were watching, their demographics and viewing
patterns, Netflix used the NoSQL database Apache Cassandras.65 Hence, Netflix executives were
aware of precisely what their customers numbering millions were viewing; they knew exactly how
famous the creations of David Fincher were, and the number of their subscribers who were Kevin
Spacey’s fans, and the number of their subscribers who had watched the original British version of
the famous and critically acclaimed House of Cards on the streaming service.
Netflix’s data signaled that Spacey or Fincher followers were also fans of House of Cards,
broadcast by the BBC in 1990. Scrutinizing the company’s data trove, Netflix managers were able
to forecast that a House of Cards version starring Spacey and directed by Fincher was the one that
Netflix viewers would be keen on viewing.66,67 Explained Steve Swasey (Swasey), Netflix’s VP of
r
IMDb or Internet Movie Database (IMDb) is a popular online database of information regarding movies,
television shows, and video games.
s
Apache Cassandra is an open source distributed management system.
7
Corporate Communications as of 2011, “We have a high degree of confidence in [House of Cards]
based on the director, the producer, and the stars. We can look at consumer data and see what the
appeal is for the director, for the stars, and for similar dramas.”68
Industry observers too appreciated Netflix for employing data about viewers’ preferences to target
them. The company had employed its viewers’ data and recommendations mechanism to
channelize viewers to the show, instead of devoting resources to promote House of Cards via
online and TV ads on other websites. Swasey added, “Through our algorithms we can determine
who might be interested in Kevin Spacey or political drama and say to them, ‘You might want to
watch this.’”69 An analyst noted that a big banner advertisement that popped up every time he
switched on Netflix and which drew his attention to the next episode of House of Cards was, to a
certain extent, responsible for his binge-viewing House of Cards.70
HOW NETFLIX BENEFITED FROM BIG DATA
According to industry observers, instead of creating shows based on viewers’ response to trial
episodes akin to the route taken by conventional stations, the video-streaming business enabled
Netflix to assess the viewing patterns and choices of its subscribers. It was aware of what the
subscribers viewed, the time at which they viewed the shows, and even when they paused to take a
recess. By incorporating the data into an algorithm, it could forecast their tastes and what would
keep them hooked.71
Observers opined that Netflix had sufficient information about the viewing choices and patterns of
its subscribers. So it chose movies and shows suitably. Explaining the advantage of Netflix basing
its decisions on data, Sarandos commented, “You get a very addressable audience. Better than that,
I know exactly who they are.”72 Netflix executives claimed that the data they had about the number
of subscribers who viewed a specific show and the number of times they viewed it helped them in
arriving at the amount they would pay to license it. Explained Sarandos, “Traditionally, content is
worth what the buyer says it is worth, but our data draws from viewer behavior to bring a bit more
science into the calculation.”73
The company was also employing the entire big data to generate content that subscribers desired to
view deploying its analytics to devise the original videos that it was financing. 74 Explained one of
the experts, “It also allows them to bid on and acquire scripts and content with insights that others
may not have…”75 Jonathan Friedland, Netflix’s communications head, explained, “We know
what people watch on Netflix and we’re able with a high degree of confidence to understand how
big a likely audience is for a given show based on people’s viewing habits. We want to continue to
have something for everybody. But as time goes on, we get better at selecting what that something
for everybody is that gets high engagement.”76
According to industry observers, the conventional studios were at a disadvantage as they clearly
had no feedback mechanisms with the movie watchers except for data regarding tickets sold and
DVDs ordered for. They were concerned with studios frittering away billions of dollars financing
shows which did not appeal to viewers. In 2011, two-thirds of the 43 new shows financed by the
large broadcast networks were withdrawn after one season, fundamentally due to poor ratings. 77 A
reputed economist’s research established that merely 22% of Hollywood movies proved to be
profitable. 78 An industry observer opined that the broadcast networks had been following the
wrong strategy by spending financial resources running into millions of dollars which could have
gone into creating better content. Netflix, on the other hand, did not have to expend huge resources
marketing its shows. It was aware of who would like its shows, and its content delivery
infrastructure could direct the targeted viewers to the concerned show. An industry insider
explained, “They’re also not spending $40 million a show on a marketing campaign. They have a
guy in a room who writes an algorithm.”79,80 Analysts pointed out that, in future, if Netflix were to
be sure that there would be audience for a show before giving it the go-ahead, it could drastically
decrease the proportion of content that would prove to be a dud. Explaining the advantage that
8
Netflix potentially had in creating original programming, given its treasure trove of big data, an
expert observed, “They’ll know when to launch a drama or a comedy or a show aimed at teens (or
tweens, or women, or…) based on data, not the hunches of some network programming head…”81
LOOKING AHEAD
Netflix was to debut four more original shows in 2013 including the shows “Hemlock Grove” and
“Arrested Development”.82 The latter would be a remake of ‘Arrested Development’, a popular
television show.83 With Netflix tasting success with House of Cards, Amazon and Microsoft were
quick off-the-block in announcing their entry into the creation of original content.84 Amazon was
also funding, on a trial basis, six comedy shows.85 Microsoft which had previously started tying up
programming deals for its Xbox Live platform, revealed that it had purchased exclusive rights to
an Indie film. 86 It was reportedly making a big foray into original content creation.87 But some
industry observers also expressed concern about Netflix stretching itself financially, given the
huge resources it had to spend on original programming (Refer to Exhibit VII for concerns
expressed in this regard).
Some experts were concerned about Netflix’s dependence on AWS for its mission-critical
operations given that during 2011 and 2012, AWS faced outage once and twice respectively.
On all these occasions, Netflix’s streaming service was down. Experts considered the last Netflix
outage which occurred on Christmas Eve to be the most damaging for the company as it happened
at a time when many Americans spent relaxed hours watching television along with their families.
Many had then opted for Netflix’s competitors to access video streaming services. 88
Some observers were also concerned that Netflix, which was intimately aware of the viewing
patterns of its subscribers, would get into trouble if it did not carefully handle the data and hence
the privacy of its customers. To substantiate their concern, they pointed out to Netflix’s previous
brush with the law. In 2006, Netflix launched the Netflix Prize, a competition inviting developers
to create algorithms that recommended movies Netflix subscribers might want to watch. However,
Netflix cancelled the competition in 2010 after some of its customers took the company to court
for infringing on their privacy. The plaintiffs claimed that the anonymous user IDs, film titles, and
ratings employed for the competition, when tallied against publicly accessible data, revealed their
identities.89 According to experts, the greater access to personal data that big data frequently
entailed would focus attention on the tension between privacy and convenience. 90
Some industry observers also foresaw a potential conflict of interest in Netflix’s deployment of big
data to create original programming as the company’s streaming traffic was majorly driven by its
recommendations of shows and movies to its viewers. They hinted at a potential controversy
emerging if Netflix’s portfolio of original programming got bigger and the company’s algorithms
diverted traffic toward its shows.91 Also, the dependence of companies such as Netflix on big data
to create original programming generated concerns over creativity. Questioned an industry
observer, “Will screenplays some day be written to meet the whims of data-driven media
streaming companies?”92 Lamented another industry observer, “If Netflix perfects the job of giving
us exactly what we want, when and how will we be exposed to things that are new and different,
the movies and TV shows we would never imagine we might like unless given the chance? Can
the auteur survive in an age when computer algorithms are the ultimate focus group?”93
Analysts expected the use of big data to grow immensely in the coming years. According to Matt
Pfeil, co-founder and VP of Customer Solutions at big data software company DataStax which
worked with Netflix to implement Apache Cassandra, “If you talk about this age as the data age,
we’re still in the teenage years, and as it matures there’s going to be orders or magnitudes of
different types of technologies that just encompass big data. The more data you have and the more
you can do with it, the smarter this business decision.”94 However, others were concerned about
the limitations of using big data. According to an industry observer, “Data can only tell you what
people have liked before, not what they don’t know they are going to like in the future.”95
9
Exhibit I
Netflix’s Financial Performance (2005-2012)
Category 2005 2006 2007 2008 2009 2010 2011 2012
Revenues 682.2 996.7 1,205.3 1,364.7 1,670.3 2,162.6 3,204.6 3,609.3
Operating Income 2.9 65.2 91.8 121.5 191.9 283.6 376.1 49.9
Net Income 42 48.9 66.7 83 115.9 160.8 226.1 17.1
Earnings per Share 0.64 0.71 0.97 1.32 1.98 2.96 4.16 0.29
(Diluted)
Net Cash Provided by 157.5 248.2 277.4 284 325.1 276.4 317.7 22.8
Operating Activities
Total Liabilities 138.4 194.6 249.2 268.3 480.6 691.9 2,426.4 3,223.2
Total Shareholders’ 226.3 414.2 429.8 347.2 199.1 290.2 642.8 744.7
Equity
* million of US$, except earnings per share (diluted)
Source: http://ir.netflix.com
Exhibit II
What is Big Data?
The term big data referred to the phenomenon of advancing trends in technology that enabled a fresh
approach to assessing the world and arriving at decisions. It represented an approach wherein decision-
making was largely dependent upon data and analysis instead of experience and intuition.96 Some experts
also employed the term ‘big data’ to describe datasets whose size was beyond the capability of normal
database software devices to capture, store, handle, and assess.97 The majority of the big data emanated
from unstructured content such as words, images, and Internet video and the surge of sensor data. The
improvements in the field of big data were largely due to the advances in the field of computing and
artificial intelligence such as natural-language processing, pattern identification, and machine learning.98
Big data was expected to generate value in five fundamental ways. First, it was expected to make
information transparent and to employ the same more often. Second, as organizations generated and saved
greater transactional information in digital mode, they were expected to garner more precise and finer
performance data on all aspects from inventories to sick leave, and hence gauge variability and bolster
performance. Third, big data enabled greater fragmentation of customers and hence delivery of more
acutely customized products/ services. Fourth, sophisticated analytics could considerably enhance
decision making, mitigate risks, and discover very important insights that would otherwise have been
unknown. Fifth, big data was expected to allow companies to evolve new products and services, improve
existing ones, and invent completely new business models.99
According to experts, big data could be employed to increase efficiencies in diverse industries. For
instance, a next-generation retailer could monitor the behavior of individual customers from their Internet
click patterns, prepare a database of their choices, and forecast their probable behavior in real time. It
would then be in a position to identify when customers were closing in on a purchase decision and prod
the sealing of the deal by bundling preferred products, combined with loyalty program savings. This real-
time targeting, which would also employ data from the retailer’s multi-layered membership rewards
scheme, would enhance purchases of merchandise with greater margins by its most valuable patrons.100
An insurance company ran though the figures about the finer details of customer behavior which enabled
it to price risk better.101 Some manufacturers employed algorithms to assess sensor data related to
production lines, developing self-regulating processes that eliminated waste, prevented expensive (and
occasionally dangerous) human interventions, and finally increased output.102
Contd…
10
Contd…
However, before going full steam ahead with the implementation of big data technologies, an
organization had to have a well-laid out plan in place. This plan had to detail the process by which the
data lying hidden in different business units or divisions was mined and integrated, the areas or business
functions in which the analyzed data would be deployed, and the tools that would enable the front-line
employees to assimilate the data into their daily operations so that the envisioned benefits were
realized.103 Every big data plan had to tackle three key challenges. First, the final benefits envisioned had
to be prioritized. Second, the organizational and the employee sensitivities had to be taken into account.
For this, simpler analytical tools had to be employed in the initial stages where the benefits derived were
modest but which would be required to boost the confidence of the employees or to take the organization
to the next level wherein more complex data would be mined and analyzed. Third, the skill-sets of
employees who were expected to deploy the data had to be developed.104
According to observers, big data could be deployed to push the envelope in the field of health care. To
illustrate this argument, they pointed out to the example of Google which had scrutinized the time and
location of online searches on flu and was able to spot probable flu outbreaks two weeks before some
national governments did.105
Compiled from various sources
Exhibit III
Netflix’s Basic Ranking Model
Source: Xavier Amatriain and Justin Basilico, “Netflix Recommendations: Beyond the 5 Stars (Part 2),”
http://techblog.netflix.com, June 20, 2012.
11
Exhibit IV
Visual Representation of the Testing Conducted by Netfllix
Source: Xavier Amatriain and Justin Basilico, “Netflix Recommendations: Beyond the 5 Stars (Part 2),”
http://techblog.netflix.com, June 20, 2012.
12
Exhibit V
Netflix’s Software Architecture
Source: Xavier Amatriain and Justin Basilico, “System Architectures for Personalization and
Recommendation,” http://techblog.netflix.com, March 27, 2013.
13
Exhibit VI
Netflix’s Cloud-based Hadoop Architecture
Source: Sriram Krishnan and Eva Tse, “Hadoop Platform as a Service in the Cloud,”
http://techblog.netflix.com, January 10, 2013.
Exhibit VII
Will Netflix Survive the Original Programming Game?
Netflix was reported to have shelled out US$40 million for 13 episodes of ‘Hemlock Grove’
which was expected to make its debut post House of Cards.106 For each episode of ‘House of
Cards’ too, Netflix was reported to have coughed up US$3.8 million, around twice the cost
generally incurred on television shows.107 Industry observers were skeptical about Netflix’s
ability to sustain the success that its first foray into original content had produced. A media
analyst pondered about the consequences if a show such as Arrested Development did not prove
to be as much a success as Netflix expected. Netflix had a lot at stake as it was ordering a
complete season or more at a single time. It would not be able to abort a show if it turned out to
be a debacle, as other networks did. 108 Also, according to observers, the entry of new rivals such
as Amazon Prime would make it tougher for Netflix to retain the same share of consumers’
discretionary income. 109
Contd…
14
Contd…
According to analysts, Netflix was shelling out more than its competitors to prevent them from
accessing shows such as ‘House of Cards’ but this would mean that it would have fewer shows
in its portfolio as its finances were limited. To maintain its leadership position, analysts opined,
Netflix was compromising its near-term profitability to invest more in original content. On
January 29, 2013, the company announced its plans to borrow $500 million to refinance around
US$225 million in debt and put more financial resources behind original programming. The
company had made commitments of over US$5 billion for the streaming content services. A
financial analyst expressed concern that Netflix’s content expenses were increasing more
speedily than its profits.110 Standard & Poor's also expressed concern over the rise in debt
leverage for Netflix. It forecast insufficient cash flows for Netflix for 2013 and probably for the
first six months of 2014 due to enhanced expenditure on original programming. S&P pointed
out that original programming entailed greater upfront payments and the returns could not be
predicted with certainty.111 An investment firm forecast that Netflix would shell out US$5
billion content costs between 2013 and 2017, with US$4.5 billion of the expenditure taking
place by 2015 itself.112 Netflix itself estimated that it would be incurring an expenditure of
US$2.1 billion on licensing content in 2013.113 However, analysts were skeptical about whether
Netflix’s original programming play at a huge cost would actually generate sufficient returns.
The company, for instance, required 520,834 more subscribers to avail of Netflix’s streaming
services for two years at a cost of US$7.99 to break even on its US$100 million investment on
House of Cards.114
Analysts were also skeptical about Netflix’s ability to compete with competitors such as
Amazon and Microsoft in the original programming space, given their financial wherewithal
(Refer to Table I for a comparison of some of the financial figures of Netflix with those of
Amazon and Microsoft for the year ended December 31, 2012).
Table I
A Comparison of Some of the Financial Figures of Netflix with those of Amazon and Microsoft
Category Netflix Amazon Microsoft
Revenues 3,609.3 61,093 73,723
Net Income (Loss) 17.1 (39) 16,978
Net Cash Provided by 22.8 4,180 31,626
Operating Activities
Total Shareholders’ Equity 744.7 8,192 66,363
* million of US$, except earnings per share (diluted)
Source: http://ir.netflix.com; http://phx.corporate-ir.net; www.micorsoft.com
Industry observers expected Amazon, particularly, to give Netflix’s executives sleepless nights.
Amazon was partly employing the considerable cash flows generated by its ecommerce business
and its capital to provide competition to Netflix’s streaming service through Amazon Prime
Video. During the third quarter of 2012, Amazon clinched a licensing deal with EPIX enabling
its users to stream movies such as The Avengers, Iron Man 2, and the Hunger Games, which
also implied that Netflix would no longer have the exclusivity that it had enjoyed with the cable
channel. 115 According to a cable executive who has spoken to Amazon, its Prime service had
also started heavily employing data-driven approaches to decide its programming. 116 And
hinting that Netflix might lose its first-mover advantage to players with deeper pockets, an
industry observer remarked, “Maybe Amazon will go beyond its tentative investments and
throw a hundred million at a different A-list series…”117 Also, Hulu, one of Netflix’s chief
competitors, was jointly owned by Fox, ABC, and CBS.118
Compiled from various sources
15
End Notes:
1
Joe Kukura, “How Big Data Stacked the Deck to Guarantee 'House of Cards' would be a Hit,” www.allvoices.com,
February 25, 2013.
2
Damian Rollison, “Big Data and Local Search: The Netflix Precedent,” http://streetfightmag.com, February 7, 2013.
3
Mike Cassidy, “Netflix “House of Cards” Offers NYT Columnist a Peek at the Real “Must See” TV,”
www.siliconbeat.com, February 25, 2013.
4
Jessica Leber, ““House of Cards” and Our Future of Algorithmic Programming,” www.technologyreview.com,
February 26, 2013.
5
“A Brief History of the Company that Revolutionized Watching of Movies and TV Shows,”
https://signup.netflix.com/MediaCenter/Timeline.
6
Joanne L. Kaufman, “Post Office Drawback Cited in Dark Forecast for Netflix,” www.nytimes.com, December 6, 2007.
7
http://investing.businessweek.com/research/stocks/snapshot/snapshot.asp?ticker=NFLX
8
Janko Roettgers, “Netflix Ends Year on a High Note, Calls House of Cards “Defining Moment for Internet TV”,”
http://paidcontent.org, January 23, 2013.
9
“Sandvine Global Report: Internet Data Usage up 120 Percent in North America,” www.sandvine.com, November
7, 2012.
10
http://investing.businessweek.com/research/stocks/snapshot/snapshot_article.asp?ticker=NFLX
11
Christopher Williams, “House of Cards: Netflix's Experiment in Binging Pays off,” www.telegraph.co.uk, February
8, 2013.
12
Andrew Leonard, “How Netflix is Turning Viewers into Puppets,” www.salon.com, February 1, 2013.
13
Michael Nixon, “House of Cards: Dealing with Big Data,” www.business2community.com, February 6, 2013.
14
“Ted Sarandos – Audience Tastes,” www.carseywolf.ucsb.edu, June 2012.
15
Adam Huttler, “Is House of Cards the Future of Cultural Programming?,” www.huffingtonpost.com, February 4, 2013.
16
Justin Sulhoff, “Netflix Uses Big Data to Produce Original Content,” https://megaplanit.com, February 13, 2013.
17
February 26, 2013.
18
Michael Cutler, “What Netflix Knows about you and why it’s a Lesson to others…,” http://cotdp.com, June 19, 2012.
19
Derrick Harris, “Netflix Analyzes a Lot of Data about your Viewing Habits,” http://gigaom.com, June 14, 2012.
20
Marshall Kirkpatrick, “Netflix's Big Data Plans to Take over the World,” http://readwrite.com, July 26, 2011.
21
Bruce Upbin, “How Intuit Uses Big Data for the Little Guy,” www.forbes.com, April 26, 2012.
22
Stephanie Overby, “In Awarding Prize for Analytics, Netflix Failed to Predict it wouldn’t be Used,” https://data-
informed.com, April 30, 2012.
23
24
Andrew Wallenstein, “Friending Facebook a Big Deal for Netflix,” http://variety.com, March 13, 2013.
25
Xavier Amatriain and Justin Basilico, “Netflix Recommendations: Beyond the 5 Stars (Part 1),”
http://techblog.netflix.com, April 6, 2012.
26
Dave Feinleib, “How the Netflix Big Data Approach is Transforming Education,” www.forbes.com, July 20, 2012.
27
28
Xavier Amatriain and Justin Basilico, “Netflix Recommendations: Beyond the 5 Stars (Part 1),”
http://techblog.netflix.com, April 6, 2012.
29
Ibid.
30
Rachel King, “Panel: Netflix, StubHub, IBM Execs Discuss Value of Big Data,” www.zdnet.com, August 10, 2012.
31
Kat Ascharya, “Media Mind: Binge Viewing and the Shifting TV Landscape,” www.mobiledia.com, October 10, 2012.
32
Ryan Lawler, “Netflix Moves into the Cloud with Amazon Web Services,” http://gigaom.com, May 7, 2010.
33
John Ciancutti, “Four Reasons we Choose Amazon’s Cloud as our Computing Platform,”
http://techblog.netflix.com, December 14, 2010.
34
Barb Darrow, “Netflix: We don’t Need No Stinkin’ Data Centers!,” http://gigaom.com, March 28, 2012.
35
“Netflix Selects Amazon Web Services to Power Mission-Critical Technology Infrastructure,” http://phx.corporate-
ir.net, May 7, 2010.
36
Joel Schectman, “Netflix Uses Big Data to Improve Streaming Video,” http://blogs.wsj.com, October 26, 2012.
37
Dan Gallagher, “Amazon Outage Shines Light on Netflix’s Big Bet,” http://blogs.marketwatch.com, December 26, 2012.
38
MG Siegler, “Netflix Original Content is Much More than a Strategy Shift — it Could Shift an Industry,”
http://techcrunch.com, March 18, 2011.
16
39
Reinhardt Krause, “Netflix's Bigger Issue than Price Hikes: Data Caps,” http://news.investors.com, July 15, 2011.
40
Alex Ben Block, “Netflix's Ted Sarandos Explains Original Content Strategy,” www.hollywoodreporter.com, April
7, 2012.
41
Steven Rosenbaum, “Netflix's Risky Strategy for 'House of Cards',” www.forbes.com, February 5, 2013.
42
7, 2012.
43
Christopher Williams, “Is Netflix about to be a Has-Been? Suddenly, Problems Seem Overwhelming,”
www.forbes.com, May 2, 2012.
44
Dean Takahashi, “Netflix Confirms Deal to Launch Kevin Spacey Series via Video Streaming,”
http://venturebeat.com, March 21, 2011.
45
7, 2012.
46
David Carr, “Giving Viewers What they Want,” http://nytimes.com, February 24, 2013.
47
Roberto Baldwin, “Netflix Gambles on Big Data to Become the HBO of Streaming,” www.wired.com, November
29, 2012.
48
Roberto Baldwin, “With House of Cards, Netflix Bets on Creative Freedom,” www.wired.com, February 1, 2013.
49
Kyle Vanhemert, “The Secret Sauce Behind Netflix’s Hit, “House of Cards”: Big Data,” www.fastcodesign.com,
February 19, 2013.
50
Thomas H. Davenport, “The Future of Entertainment is Analytical,” http://mobile.blogs.wsj.com, March 6, 2013.
51
29, 2012.
52
Nellie Andreeva, “Netflix to Enter Original Programming with Mega Deal for David Fincher-Kevin Spacey Series
‘House of Cards’,” www.deadline.com, March 15, 2011.
53
Michael Berliner, “Internet TV Commissioning Round-Up,” www.guardian.co.uk, February 21, 2013.
54
Cliff Edwards, “Netflix's Big Gamble with 'House of Cards',” www.sfgate.com, February 4, 2013.
55
Greg Satell, “What Netflix's 'House of Cards' Means for the Future of TV,” www.forbes.com, March 4, 2013.
56
Dieter Bohn, “House of Cards the 'Most-Watched' Show on Netflix, will 'Arrested Development' Follow?,”
www.theverge.com, February 12, 2013.
57
Brian Stelter, “Release of 13 Episodes Redefines Spoiler Alert,” www.nytimes.com, February 5, 2013.
58
“'House of Cards' Getting Thousands of Mentions per Hour on Social Media,” www.allvoices.com, February 2, 2013.
59
“'House Of Cards': A Sign that Netflix can Become a Content Creator,” http://seekingalpha.com, March 13, 2013.
60
Natan Edelsburg, “Is ‘House of Cards’ a Success? Social TV has the Answer,” http://lostremote.com, February 18, 2013.
61
Laura Hudson, “The Poll Results are in: You’re Totally Binge-Watching House of Cards,” www.wired.com,
February 5, 2013.
62
Luke Westaway, “House of Cards is Most-Watched Show on Netflix, Everywhere,” http://crave.cnet.co.uk,
February 13, 2013.
63
Dieter Bohn, “House of Cards the 'Most-Watched' Show on Netflix, will 'Arrested Development' Follow?,”
www.theverge.com, February 12, 2013.
64
“'House Of Cards': A Sign that Netflix can Become a Content Creator,” http://seekingalpha.com, March 13, 2013.
65
Sophie Curtis, “Netflix Foretells 'House of Cards' Success with Cassandra Big Data Engine,” Techworld, March 28, 2013.
66
“The 'Big Data' Revolution: How Number Crunchers Can Predict our Lives,” www.npr.org, March 7, 2013.
67
68
Ryan Lawler, “How Netflix Will Use Big Data to Push House of Cards,” http://gigaom.com, March 18, 2011.
69
Ibid.
70
Daniel Ferry, “A Bear Gives a Bull Case for Netflix,” www.fool.com, February 24, 2013.
71
Margaret Rock, “The Year Ahead: Finding Wisdom in Big Data,” www.mobiledia.com, December 17, 2012.
72
Jason Lange, “Netflix, Big Data, and What Hollywood is Missing out on.,” www.jasonlange.me, September 24, 2012.
73
74
Peter Kafka, “Big Data, Soft Sell: Netflix Pitches a Hands-Off Approach to Hollywood,” http://allthingsd.com,
March 1, 2013.
75
Jesse Hirsh, “Netflix, Big Data, and House of Cards,” http://jessehirsh.ca, February 5, 2013.
76
29, 2012.
77
17
78
Thomas H. Davenport, “The Future of Entertainment is Analytical,” http://mobile.blogs.wsj.com, March 6, 2013.
79
Andrew Wallenstein, “Netflix Series Spending Revealed,” http://variety.com, March 8, 2013.
80
81
Rick Liebling, “The Media House of Cards and Netflix’s Big Disruption,” www.fastcocreate.com, February 28, 2013.
82
83
Peter Pham, “The Future of Netflix is Data Driven,” http://beta.fool.com, March 9, 2013.
84
Greg Satell, “5 Trends that Will Drive the Future of Technology,” www.forbes.com, March 12, 2013.
85
Brian Stelter, “Release of 13 Episodes Redefines Spoiler Alert,” www.nytimes.com, February 5, 2013.
86
Greg Satell, “What Netflix's 'House of Cards' Means for the Future of TV,” www.forbes.com, March 4, 2013.
87
Andrew Wallenstein, “Netflix Series Spending Revealed,” http://variety.com, March 8, 2013.
88
Joel Schectman, “Netflix Amazon Outage Shows ‘any Company can Fail’,” http://blogs.wsj.com, December 27, 2012.
89
Clint Boulton, “CIOs must Keep Big Data Private,” http://mobile.blogs.wsj.com, October 23, 2012.
90
Brad Brown, Michael Chui and James Manyika, “Are you Ready for the Era of ‘Big Data’?,”
www.mckinseyquarterly.com, October 2011.
91
92
February 26, 2013.
93
94
Sophie Curtis, “Netflix Foretells 'House of Cards' Success with Cassandra Big Data Engine,” Techworld, March 28, 2013.
95
96
Steve Lohr, “The Age of Big Data,” www.nytimes.com, February 11, 2012.
97
James Manyika, Michael Chui, Brad Brown, Jacques Bughin, Richard Dobbs, Charles Roxburgh and Angela Hung Byers,
“Big data: The Next Frontier for Innovation, Competition, and Productivity,” www.mckinsey.com, May 2011.
98
Steve Lohr, “The Age of Big Data,” www.nytimes.com, February 11, 2012.
99
James Manyika, Michael Chui, Brad Brown, Jacques Bughin, Richard Dobbs, Charles Roxburgh and Angela Hung Byers,
“Big data: The Next Frontier for Innovation, Competition, and Productivity,” www.mckinsey.com, May 2011.
100
101
Jacques Bughin, John Livingston and Sam Marwaha, “Seizing the potential of ‘big data’,”
102
103
Stefan Biesdorf, David Court and Paul Willmott, “Big Data: What’s your Plan?,” www.mckinseyquarterly.com,
March 2013.
104
Ibid.
105
Nayan Chanda, “Can Big Data Save Lives?,” www.businessworld.in, June 22, 2012.
106
Kevin Fitzpatrick, “How Much Money is Netflix Putting into ‘Hemlock Grove?’,” http://screencrush.com, June 12, 2012.
107
Roberto Baldwin, “With House of Cards, Netflix Bets on Creative Freedom,” www.wired.com, February 1, 2013.
108
29, 2012.
109
“Analysis: Why Netflix must Rethink Binge Viewing,” http://variety.com, January 29, 2013.
110
111
Debbie Cai, “S&P Lowers Netflix Outlook to Negative on Higher Debt Leverage,” http://online.wsj.com, January
29, 2013.
112
Tom Cheredar, “What the Disney Streaming Deal Means for Netflix’s Future,” http://venturebeat.com, December 5, 2012.
113
Tom Cheredar, “Netflix’s Reed Hastings Says Amazon Prime Loses $1B Annually,” http://venturebeat.com,
November 17, 2012.
114
Rebecca Greenfield, “The Economics of Netflix's $100 Million New Show,” www.theatlanticwire.com, February 1, 2013.
115
Agustino Fontevecchia, “Amazon Eyeing Everyone's Lunch: Bezos Going after Apple, Netflix, Big Data, and
Cloud Computing,” www.forbes.com, October 25, 2012.
116
117
Tim Wu, ““House of Cards” and the Decline of Cable,” www.newyorker.com, February 4, 2013.
118
18

Module 3. Netflix - Leveraging Big Data

Uploaded by

Copyright:

Available Formats

Module 3. Netflix - Leveraging Big Data

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Module 3. Netflix - Leveraging Big Data

Uploaded by

Copyright:

Available Formats

ITSY/075

IBS Center for Management Research

Netflix: Leveraging Big Data to Predict Entertainment Hits

 2013, IBS Center for Management Research. All rights reserved.

Netflix: Leveraging Big Data to Predict Entertainment Hits

BIG DATA AT NETFLIX

SHIFT FROM OWN DATA CENTERS TO AMAZON WEB SERVICES

THE NEED FOR ORIGINAL CONTENT AT NETFLIX

ENTER THE ‘HOUSE OF CARDS’

IDENTIFYING THE SHOW

HOW NETFLIX BENEFITED FROM BIG DATA

Compiled from various sources

Netflix’s Basic Ranking Model

Visual Representation of the Testing Conducted by Netfllix

Netflix’s Software Architecture

Netflix’s Cloud-based Hadoop Architecture

You might also like