Analyzing the ways IT has changed our TV consumption: Binge Watching and Marathon Watching

Conference Paper (PDF Available) · December 2017with 842 Reads 
How we measure 'reads'
A 'read' is counted each time someone views a publication summary (such as the title, abstract, and list of authors), clicks on a figure, or views or downloads the full-text. Learn more
Conference: International Conference on Information Systems, At Seoul, South Korea, Volume: 38th
Cite this publication
Abstract
Rather than viewing television shows one episode at a time, many people now consume them back-to-back. While this is in itself a unique practice, it manifests in various forms, two of which seem to enjoy great popularity: ‘bingeing’ and ‘marathoning’. In this study, we explore their association with online television streaming services and scheduled television. Further, we examine whether these technologies play a role in people’s consumption decisions in terms of content quality, an issue, which has so far received no academic attention. This we believe is of great importance to both consumers and content providers. We analyze the relationship based on user-generated data extracted from Twitter. The findings reveal that individual viewers consume higher quality content while bingeing than while marathoning. We discuss the implications of our results for theory and practice.
Figures - uploaded by Jani Merikivi
Author content
All content in this area was uploaded by Jani Merikivi
Content may be subject to copyright.
Analyzing the ways IT has changed our TV consumption:
Binge Watching and Marathon Watching
Journal:
International Conference on Information Systems 2017
Manuscript ID
ICIS-1060-2017.R1
Track:
13. Human behavior and IS
Keywords:
binge watching, marathon viewing, online television streaming services,
system usage, digital content quality, digital consumpti, binge watching,
marathon viewing, online television streaming services, system usage,
digital content quality, digital consumption decision-making, user-
generated data, behavioral manifestations and outcomes
Abstract:
Rather than viewing television shows one episode at a time, many people
now consume them back-to-back. While this is in itself a unique practice, it
manifests in various forms, two of which seem to enjoy great popularity:
‘bingeing’ and ‘marathoning’. In this study, we explore their association
with online television streaming services and scheduled television. Further,
we examine whether these technologies play a role in people’s
consumption decisions in terms of content quality, an issue, which has so
far received no academic attention. This we believe is of great importance
to both consumers and content providers. We analyze the relationship
based on user-generated data extracted from Twitter. The findings reveal
that individual viewers consume higher quality content while bingeing than
while marathoning. We discuss the implications of our results for theory
and practice.
Analyzing binge watching and marathon watching
Thirty
eigh
th International Conference on Information Systems, Seoul 2017 1
Analyzing the ways IT has changed our TV
consumption: Binge Watching and Marathon
Watching
Completed Research Paper
Bikesh Raj Upreti
Department of Information and
Service Economy, Aalto University
School of Business, Runeberginkatu
22-24, Helsinki, FI 00100, Finland
bikesh.upreti@aalto.fi
Jani Merikivi
Department of Information and
Service Economy, Aalto University
School of Business, Runeberginkatu
22-24, Helsinki, FI 00100, Finland
jani.merikivi@aalto.fi
Johanna Bragge
Department of Information and
Service Economy, Aalto University
School of Business, Runeberginkatu
22-24, Helsinki, FI 00100, Finland
johanna.bragge@aalto.fi
Pekka Malo
Department of Information and
Service Economy, Aalto University
School of Business, Runeberginkatu
22-24, Helsinki, FI 00100, Finland
pekka.malo@aalto.fi
Abstract
Rather than viewing television shows one episode at a time, many people now consume
them back-to-back. While this is in itself a unique practice, it manifests in various forms,
two of which seem to enjoy great popularity: ‘bingeing’ and ‘marathoning’. In this
study, we explore their association with online television streaming services and
scheduled television. Further, we examine whether these technologies play a role in
people’s consumption decisions in terms of content quality, an issue, which has so far
received no academic attention. This we believe is of great importance to both
consumers and content providers. We analyze the relationship based on user-generated
data extracted from Twitter. The findings reveal that individual viewers consume
higher quality content while bingeing than while marathoning. We discuss the
implications of our results for theory and practice.
Keywords: binge watching, marathon viewing, online television streaming services, system
usage, digital content quality, digital consumption decision-making, user-generated data, behavioral
manifestations and outcomes
Page 1 of 16
Analyzing binge watching and marathon watching
Thirty
eigh
th International Conference on Information Systems, Seoul 2017 2
Introduction
For a television to bring value to its users in a world of mostly free media, one would intuitively assume
that it must offer enticement to viewing. When it does, users will exhibit an autonomous and selective
orientation towards their preferences. Rust and Alpert (1984) present this viewing choice as a two-stage
process. In stage one, users select whether they watch television at all, and, in stage two, determine, which
television show they wish to follow. The assumption follows uses and gratifications theory (Katz 1959;
Katz et al. 1974), according to which users select the medium and content that satisfies their needs
(Ruggiero 2000). The users represent active consumers, who have control over their viewing: they decide
whether television delivers content, which brings them value, and, then, will consistently act accordingly
(McQuail et al., 1972). Under this theory, Webster (2014), for example, admits that knowing users’
preferences would certainly help explain what television shows they view. Yet, he reminds that viewing
choices are less straightforward, as they depend on industry structures like competition and program
scheduling (Goodhardt et al. 1987; Webster and Newton 1988).
In this study, we follow Webster and Newton (1988) and recognize both the industry structures and user
preferences. We assume that audience availability (e.g., Pingree et al. 2001), that is, the number of people
who are likely to use television at any point in time will remain relevant even when television has gone
online. Breaking away from program scheduling has no doubt made it easier to practice their preferences,
yet availability is peaking during the evenings (De Feijter et al. 2016; Pullar-Strecker 2015). This is
because users must still fit their television viewing around everyday commitments (e.g., family and work).
We also assume that viewing television during the evening hours goes beyond a mere habit; that users’
relationship with television springs from a personal preference for a certain television show rather than
the medium itself. To illustrate this, we focus on users who realize this preference for a television show by
consuming its episodes in bulk. The practice is not different from anything seen or known before. Indeed,
we have witnessed users have engaged in occasional television marathons at least for three decades, and
will likely continue to do so. When HBO, for example, celebrated the 2016 holiday season by airing one
season of Game of Thrones each day for six consecutive days, not only did new users join in to catch up on
the show (the last season premiered six months later), but also the old viewers who wanted to refresh
their memories and fan connections.
Today, thanks to online television streaming services (e.g., Amazon Video, Hulu, and Netflix), users are
able to express their preference for a certain television show easier than ever before. They no longer have
to wait for marathons or pile up episodes of their favorite shows onto video tapes, DVDs, or set-top boxes.
With unrestricted access and use of content, viewers will miss no episode due to scheduling conflicts, and
most importantly, they can start and end their sessions at their own convenience. When these sessions
revolve around a certain television show and extend beyond one episode, they develop into binge
watching – a popular cousin to marathoning (Harris Interactive 2013; Netflix 2013; Nielsen 2013).
What distinguishes marathoning from bingeing is that it is less autonomous. Users must engage in
marathons at a predetermined time and the television shows they consume are preselected. This leaves
them with a Hobson’s choice
1
: the users may select whether they decide to participate in a marathon, but
they cannot select among television shows nor the time they wish to start and end their sessions. In this
study, we are interested in the implications bingeing and marathoning have. For example, does bingeing
differ from marathoning in terms of content? If yes, how well these two practices correlate with viewing
ratings? Moreover, does autonomy narrow the bingers’ taste for television shows?
As of today, no researcher has undertaken a study on bingeing and marathoning to scrutinize their link to
the IS artifact. The focus of past research has been on user motives (Pittman and Sheehan 2015), health
issues (Devasagayam 2014), and advertising (Schweidel and Moe 2016). The timely question of how
scheduled television and online television streaming services influence television viewing has thus
remained unanswered. Therefore, we focus on two system specific viewing manifestations: ‘bingeing’ and
‘marathoning’. The former occurs via online television streaming technologies, whereas the latter takes
1
Thomas Hobson ran a carrier and horse rental business in Cambridge, England around the 17
th
century. He
became famous for limiting the choice his customers could make among the horses they wanted to rent. He obliged
them to take the horse, which stood next to the stable door. That is, he refused to hire out horses other than the ones
he chose (“this or none”). Therefore, the choice his customers made refers to Hobson’s choice.
Page 2 of 16
Analyzing binge watching and marathon watching
Thirty
eigh
th International Conference on Information Systems, Seoul 2017 3
place via scheduled television. We connect these manifestations with two distinct elements, time and
content, which we assume help us reveal the system-specific outcomes.
We extract data from a popular social networking service, Twitter. We select Twitter, as it is one of the key
“digitized water coolers”, through which users discuss television shows (Matrix 2014). We believe these
unstructured “voice of the market” discussions (Chen et al. 2012, 1169) allow us to map between the
selected viewing manifestations and the technologies, which makes these manifestations possible. To
analyze whether the technologies influence the content viewers consume, we complement the Twitter data
with user-generated ratings from IMDb.
By taking these steps, we believe our paper will lay a groundwork for future research in two ways. First,
there is relatively little research on binge watching and marathon watching, and their outcomes. Such
research is crucial as it helps us to identify the differences between binge watching and marathon
watching. It also reminds that both individual and structural determinants are relevant when modeling
system use. Second, by focusing on online television streaming services and scheduled television, which
offer these two manifestations (binge watching and marathon watching), we shed light on how user
autonomy influences users’ attempts to bend these systems to their own purposes (Friedman 1996).
The remainder of this paper proceeds as follows. Next, we provide a review of related research. By this, we
seek to establish that bingeing and marathoning are distinct manifestations of active television viewing.
We then describe our data, which we use to verify these manifestations. We also test whether these
manifestations differ in two dimensions: content selection and time use. Finally, we present our results
and conclude with a discussion of implications and directions for future work.
Related Research
Bingeing vs. Marathoning
As per uses and gratifications theory (Katz et al. 1974), we assume that users watch television shows that
best satisfy their preferences. We draw support for our assumption from users who watch episodes of a
certain television show back-to-back (Schweidel and Moe 2016). Here, users are engaged in a television
show, not television itself. The television show captivates their attention to the extent they prefer to
continue even after the scheduled episode has already ended. Under this assumption, viewing a specific
television show is a primary activity, during which users are attempting to make sense of what happens on
television at a given moment. That is, users interact with the television show they are viewing: they form
opinions about the show, perceive emotions in it, or talk about it with other people (see Glebatis Perks
2014). This makes that the engagement in television shows can, but does not have to, be introverted.
Giving form to one’s thoughts or emotions may also promote social interaction (Jenner 2015). Therefore,
flow (Csikszentmihalyi 1991) or cognitive absorption (Agarwal and Karahanna 2000), is certainly one
aspect, but not the whole picture of user engagement.
To offer possibilities for engagement in television shows, television networks depart occasionally from
their daily, weekly, or monthly schedules and host television marathons. While these marathons allow
users to consume full seasons or even complete television shows, the users cannot choose the day or time
when marathons start and end. This means they would miss episodes if they did not complete the
marathons. Perhaps even more inconvenient, the users cannot choose what television show they wish to
marathon view. To overcome these hurdles, users could record television shows and view them after live
broadcasting. This way, they may choose when to watch television shows without missing any episodes –
even if they watched them at their own pace. Still, their selections depend on program scheduling
(Hobson’s choice). In order to gain greater control over time and content, users may turn to online
television streaming services, as they let users to decide when and what television shows they want to
watch one episode after another.
Given the above, it seems clear that while these technologies allow users to watch several episodes of the
same television show back-to-back, their support for autonomy varies: scheduled television is less
autonomous than online television streaming services (and time shifting). Therefore, we distinguish
between two system-specific viewing manifestations, binge watching and marathon watching (Figure 1),
and explore them from two perspectives: content selection and time use. Binge watching departs from
program scheduling. It took off in the 1990s when television shows became available on box sets
Page 3 of 16
Analyzing binge watching and marathon watching
Thirty
eigh
th International Conference on Information Systems, Seoul 2017 4
comprising of full seasons for extended viewing (Graves 2015). But it did not turn into a buzzword until
Netflix, a popular online television streaming service, started releasing episodes in bulk as a response to
user preferences and the new norms of viewer control’ (Matrix 2014, 133). Marathon watching, in turn,
has never left scheduled television. It still taps into extended time slots to indulge users in one particular
television show (Kompare 2002).
Figure 1. Differences between Binge Watching and Marathon Watching
User Autonomy: Content Selection and Time Use
We seek to cover two distinct outcomes, which we assume result from both binge watching and marathon
watching: content selection and time use. As discussed earlier, these behavioral outcomes depend on user
autonomy, implying it can be determined either by the system or by the user. By user autonomy, we refer
to users’ ability to act on their own reasons, desires, and goals (Friedman 1996). That is, users must fully
endorse using the system. Following self-determination theory, user autonomy does not rule out external
inputs provided the users are wholeheartedly consent to them (Ryan and Deci 2006). It is thus possible
that users discover desires through system use or that other users influence their desires.
Why autonomous use is a powerful element is because it contributes to satisfying users’ desires.
According to Friedman (1996), for a system to facilitate user autonomy, it should offer its users the
necessary capabilities to realize their desires. These capabilities concern mastering functions, which link
to higher order desires and goals, such as booking a flight ticket or buying stocks online. In our case,
television users would certainly appreciate user autonomy over content selection and time use.
Time use relates to users’ ability to satisfy their desires at their own convenience. We expect that users of
online television streaming services have a genuine opportunity to schedule their bingeing to their liking
and around their existing commitments. This user autonomy over time we believe is likely to scatter binge
watching sessions across time. When more pressing matters emerge, it is at least possible for the users to
reschedule their binge watching as they know they can continue viewing at a time more convenient to
them (Wortham 2011). Alternatively, when users marathon-watch television shows on scheduled
television, they cannot decide on what day and at what time television marathons occur. Typically,
marathons take place during holiday seasons or prior to season premieres (Conlin 2015; Hewitt 2015). If
users do not hop in these marathons, or they hop out of them before the end, they will miss them either
completely or partially. Further, knowing that television marathons are a onetime opportunity makes
decision-making even more difficult. If users were afraid of missing the opportunity that other people
seize, it would motivate them to consume television shows without expecting too much from their quality
(Conlin et al. 2016). A system, which offers its users less autonomy over time, would thus bring them less
confidence in content evaluation.
This leads us to the second outcome element, content, which concerns television shows users consume.
Content is of crucial importance for binge watching, a claim that derives support from Jenner (2015) and
Matrix (2014), both of whom associate binge watching to consuming television shows of high quality.
Page 4 of 16
Analyzing binge watching and marathon watching
Thirty
eigh
th International Conference on Information Systems, Seoul 2017 5
Matrix (2014) specifies that quality expectations relate to Netflix in particular. Jenner (2015, 1) adds that
television shows, which are ‘binge-worthy’ – a reflection of high quality – justify “the extra expense of
buying the DVDs or the VOD subscription”. This, however, may not be the case with marathon watching,
as users face the Hobson’s choice. They are simply beyond being able to maximize their utility because
there is only one television show they can choose. Yet, this does not have to mean that the television show
is always of poor quality. The users just cannot make a comparison between numerous television shows,
implying that they merely express a preference for marathon watching, not for a particular television
show.
When using online television streaming services, users get to choose among a variety of television shows.
Drawing upon rational choice theory, they will then select the best possible television show they want to
consume. Given that they can also binge watch their selection at their own convenience, we believe that
the more autonomous the users are over their content the better the television show fits to their liking
(Green and Srinivasan 1978). This, in turn, makes us expect that binge watching correlates more with
higher ratings than marathon watching. We present our research framework in Figure 2.
Figure 2. Research Framework
Research Method
Data Collection
Due to lack of access to actual log data (Vander Verff 2016), we rely on one of the most popular social
networking service, Twitter, and collect data on its users’ viewing practices. In Twitter, users can read,
share and post short 140-character tweets, which reveal information on their activities, preferences, and
opinions. As such, it provides a useful source for user generated data that are publicly available real-time.
While we had no full access to its large user base of nearly 320 million active users who write over 500
million tweets every day (Aslam 2017; Statista 2017), Twitter still allows drawing a significant data sample
that goes beyond a typical survey sample. More than the sample size, however, we are interested in its
ability to provide ‘on-the-spot’ viewing data across multiple television networks and online television
streaming services. In addition, it allows specifying the data contained in each message with a set of
‘binge’ and ‘marathon’ related keywords.
In order to extract the tweets relevant to our study, we connected to a streaming endpoint to query for
data. The query was open between October 30, 2014 and February 20, 2015 (113 days). For this, we used a
streaming application-programming interface (streaming API) through which we requested the tweets
from Twitter. Since Twitter poses a limit to queries if the tweets that match the criteria exceed one percent
Page 5 of 16
Analyzing binge watching and marathon watching
Thirty
eigh
th International Conference on Information Systems, Seoul 2017 6
of all currently posted tweets, not all tweets are necessarily accessible (see also Morstatter et al. 2013).
However, given its users send an average of 350,000 tweets per minute, we believe this was not an issue.
Once having created the connection, we downloaded the requested tweets into a NoSQL database
(MongoDB).
Cleaning and Preprocessing
We collected 661,825 tweets (see Appendix A1). Of these, we removed the duplicates the streaming API
created due to co-existence of multiple keywords. We also removed the user-created duplicates, that is, all
tweets and re-tweets (i.e., re-posted tweets) that contain exactly the same text from the same user. We did
this, as we wanted each tweet to indicate a unique viewing activity.
2
Next, we deleted tweets, which were
either advertisements, marketing, promotions, or contained web links. These actions reduced our data to
134,938 tweets.
Before analyzing the data, we preprocessed the tweets. This helped us to exclude irrelevant tweets, to
eliminate unnecessary details, and to reveal important textual features. However, revealing these
important features from the tweets is a challenging lingual process. This is because the tweets manifest
informal communication, typically including misspellings, colloquial language, punctuation errors,
unusual abbreviations, and poor grammatical structure. Further, relevant terms and concepts may hide in
long hashtags that combine multiple terms (for example, #teenwolfmarathon). The tweets could also
contain usernames, hiding the keywords (i.e., ‘binge’, ‘marathon’, or one of their many variants). To
complicate the process even more, our keywords may also concern off topic activities and events, such as
binge eating, movie binges, and running events. Marathon alone had over 600 open term compounds
(e.g., marathon run, New York marathon, and marathon classic), whereas binge related to 500 open term
compounds (e.g., binge eating, binge smoke, and House of Cards binge). Relying blindly on the keywords
would thus have compromised our sample. Therefore, we decided to confirm the tweets by using a
television show title besides our keywords and their variants.
For this, we tokenized the terms ‘binge’ and ‘marathon’ and isolated them from compound words and
hashtags. To identify the appropriate television show titles, we derived a list of over 2,400 titles from
IMDb
3
. We then used the list to map the titles onto their colloquial equivalents, abbreviations, acronyms,
hashtags, and misspellings, that is, those that appear in the tweets. For example, to link ‘House of Cards’,
a popular television show produced by Netflix, we searched for terms like ‘hoc’, ‘hocs’, ‘housecards’, and
‘frank underwood’ (name of the protagonist). Further, given many of the titles are compound terms, we
replaced them with one single term. To avoid vague and incorrect term replacement, we carried out exact
and partial string matching (Sanchez 2013) using two separate mapping lists for each matching scheme.
After having mapped the titles, we drew a subset of relevant tweets.
Lastly, we limited our analysis to popular television shows as their popularity decides system success.
Therefore, we excluded rare television shows, which attracted only occasional viewers. We then extracted
a list of popular television shows with at least 100 unique tweets, a threshold equivalent to one tweet per
day. Of these television shows, we decided to exclude those producing unsolved ambiguity (such as
‘Friends’). This led to a list of 186 television shows.
Model Specification
Our analysis draws on counting tweets. We started by classifying the tweets into two categories depending
on whether it is the term ‘binge’ or ‘marathon’ that appears along with a television show title. We then
computed how many times each television show was tweeted in binge and marathon categories. Next, we
compared between binge watching and marathon watching. For this, we used a selection of television
show attributes (see below). We estimated two separate models; one with binge tweet counts and another
with marathon tweet counts (count as dependent variable). Given that the dependent variables are counts,
2
At the time we collected the data, Twitter had no feature that allowed adding comments to re-tweets. Therefore,
users cannot indicate retweet “I am binge watching Breaking Bad” and add, for example, “I am too.”
3
A comprehensive list of 2450 television show titles provided by a user called “International-Television-Award”.
http://www.imdb.com/list/ls009728319/
Page 6 of 16
Analyzing binge watching and marathon watching
Thirty
eigh
th International Conference on Information Systems, Seoul 2017 7
we employed a generalized linear model (GLM) with negative binomial distribution (O'Hara and Kotze
2010). We found that the tweet frequency skewed towards a few television shows, which claimed a large
number of tweets, while the majority of the television shows had only a moderate tweet frequency. Thus,
the use of negative binomial distribution as an alternative to Poisson distribution allowed us to better
account for the large variance in the tweet frequencies.
To describe the television shows, we used the following attributes: ratings, genre, release date, continuity,
number of seasons, and number of episodes. Glebatis Perks (2014), states that users settle in for
marathons because they appreciate the quality these marathons deliver. In a similar vein, Jenner (2015)
points out that when binge watching users seem to deliberately choose to view television shows of high
quality. To capture the content quality, we used the title ratings. Further, we used genre as a measure of
content diversity since we wanted to find out whether the selected system makes users narrow down their
viewing options. Release date informs us about users’ preferences between ongoing and already finished
television shows. While we acknowledge that people care for both old and new television shows, it is likely
they choose to consume content that they have not yet seen. Hence, we believe users prefer the more
recent arrived content (Jenner 2015). Further, we control television show attributes, such as the number
of seasons and number of episodes, which are indicative of the time viewers spend on their binges and
marathons. To gather these data, we used IMDb, a frequently visited movie and television show website,
as the primary database, and Wikipedia as a complementary data source. Table 1 provides the complete
list of our variables.
To estimate the two models, we opted for the Bayesian approach as it allows us to estimate credible
intervals for the quantities of interest. The advantage of these intervals follows from the fact that credible
interval estimates draw on the most probable interval of observing the model parameter. Further, in the
experiment with small sample size (186 television shows), the Bayesian estimation of credible interval
provides an edge over frequentist interval estimates (Chen and Chao 1999). For instance, a 95 percent
credible interval [q
1
, q
2
] for a variable means that there is 0.95 post-data probability that the value of the
parameter lies in this region (i.e., the parameter is greater than q
1
but less than q
2
). This is subtly different
from (frequentist) confidence intervals, which one interprets only in relation to a sequence of similarly
repeated experiments (Gelman et al. 2014). Recently in applied statistics, there has been increasing
emphasis on interval estimation rather than hypothesis testing, providing a strong motivation for the use
of Bayesian models (Gelman et al. 2014; Johnson 1999). In our paper, we estimate the credible intervals
by running Markov chain Monte Carlo (MCMC) simulation with 20,000 samples (50% warm-up), 10
chains and uninformative prior using an R package “rstanarm”
4
.
Table 1. Description of the Model Variables
Variable Title Description Type
Marathon Tweet #
Number of Tweets where TV show title appears with marathon term Count
Binge Tweet # Number of Tweets where TV show title appears with binge term Count
IMDb Rating IMDb users’ rating of the TV show Integer
Starting Year 2014 - Year in which the TV show started (age of the show, base 2014)
Numeric
Continuing Is the TV show still continuing or has it already ended (base 2014) Binary
No. of Seasons Number of seasons for the TV show (before Feb 21, 2014) Numeric
No. of Episodes Number of episodes for the TV show (before Feb 21, 2014) Numeric
Genre A TV show belongs to 1 - 3 genres (out of 20 alternatives) Binary
4
Stan Development Team (2016). rstanarm: Bayesian applied regression modeling via Stan. R package
version 2.13.1. http://mc-stan.org/.
Page 7 of 16
Analyzing binge watching and marathon watching
Thirty
eigh
th International Conference on Information Systems, Seoul 2017 8
Results
Tweet Dynamics
Figure 3 illustrates the distribution of both binge watching and marathon watching related tweets. Apart
from the exceptional peak in February when Netflix released the second season of House of Cards,
5
binge
watching fluctuates less than marathon watching (tweets per day). Marathon watching seems to peak
during holiday seasons, implying that television networks seek to schedule their marathons for occasions
when audience availability is high. Going by day of week, binge watching also spreads more steadily over
the week. The peak on Thursday reflects the peak after the House of Cards release date. The results
comply with our expectation that binge watching is associated with greater user autonomy.
To verify our results, we examined the dates for television marathons. The networks’ program schedules
confirm these peaks concerning the marathon watching.
Tweet Frequencies per Day Tweet Frequencies per Day of Week
Figure 3. Tweet Frequencies Per Day and Per Day of Week
Top Television Shows
Concerning binge watching, the results show that Netflix’s House of cards was the most tweeted television
show (12,073), leaving Breaking Bad (2,709) and Dexter (2,456) far behind. As for marathon watching,
The walking dead’ was leading with 8,533 tweets. The television shows that were at the top in the binge-
watching category either had ended very recently or were still running (at the time of data collection). This
points towards the argument that online television streaming service users binge watch trending
television shows. Another interesting notion is that the top television shows in both categories are mainly
drama, thriller, or crime. While the ratings imply that the television shows are of rather high quality, some
shows (e.g. ‘Teen Wolf’) have a rating below eight (7.7).
5
Surprisingly, the tweets on House of Cards peaked no earlier than six days after Netflix released the
second season. Netflix released the second season on Valentine’s Day (Feb 14, 2014), which is not a public
holiday (Friday). Still, it would have been more natural that either the tweets peaked on the release date
or during the following three-day weekend (President’s Day was on Feb 19, 2014) (Wallenstein 2014).
2013-11-01
2013-11-05
2013-11-09
2013-11-13
2013-11-17
2013-11-21
2013-11-25
2013-11-29
2013-12-03
2013-12-07
2013-12-11
2013-12-15
2013-12-19
2013-12-23
2013-12-27
2013-12-31
2014-01-04
2014-01-08
2014-01-12
2014-01-16
2014-01-20
2014-01-24
2014-01-28
2014-02-01
2014-02-05
2014-02-09
2014-02-13
2014-02-17
0
2k
4k
6k
8k
1
0
k
Number of Tweets
Marathon Binge
Sunday Monday Tuesday Wednesday Thursday Friday Saturday
0
2k
4k
6k
8k
10k
12k
14k
N
u
m
b
e
r
o
f
T
w
e
e
t
s
Marathon Binge
Page 8 of 16
Analyzing binge watching and marathon watching
Thirty
eigh
th International Conference on Information Systems, Seoul 2017 9
Table 2. Top Marathoned or Binged TV Shows Based on Tweet Frequency
No. of
tweets TV show
IMDB
rating
(1-10)
Start
year
End
year*
No. of
seasons/
episodes
Genres
Original
show
provider
Top-5 television shows (marathon watching)
8,533 The Walking Dead 8.8 2010 2014* 4 / 45 Drama; Horror; Thriller AMC
6,510 Law & Order 7.6 1990 2010 20 / 456 Crime; Drama; Mystery NBC
6,231 Breaking Bad 9.6 2008 2013 5 / 62 Crime; Drama; Thriller AMC
3,604 Teen Wolf 7.7 2011 2014* 3 / 43 Action; Comedy; Drama MTV
3,417 Pretty Little Liars 8.1 2010 2014* 7 / 91 Drama; Mystery; Romance ABC
Top-5 television shows (binge watching)
12,073 House of Cards 9 2013 2014* 2 / 26 Drama Netflix
2,709 Breaking Bad 9.6 2008 2013 5 / 62 Crime; Drama; Thriller AMC
2,456 Dexter 9.1 2006 2013 9 / 96 Crime; Drama; Mystery Showtime
2,179 Scandal 7.9 2012 2014* 3 / 47 Drama; Thriller ABC
1,759 Sherlock 9.3 2010 2014* 3 / 9 Crime; Drama; Mystery BBC
* Ongoing television shows
Term Co-Occurrence Analysis
To verify that binge watching links to online television streaming services and marathon watching links to
scheduled television, we conducted a co-occurrence analysis. It draws on the co-occurrence between the
20 most tweeted television shows (in our sample) and their sources (a total of seven television networks
and online television streaming services). In addition, since both ‘binge’ and ‘marathon’ are components
of multi-word terms (e.g., ‘binge watch’ or ‘marathon view’), we examined their association with the most
typical parent verbs: ‘watch’ and ‘view’. To represent the results, we visualized the co-occurrences (i.e.,
behavior + television show + source) and their relationships using pairwise conditional probabilities
(PCP). That is, we computed conditional pairwise probabilities for all the observed term co-occurrences.
For visual clarity, we limited the edges between the terms to a minimum of five (5) percent co-occurrence.
The thickness of arrows depicts the strength of the relationship between the constructs: the thicker the
arrow the stronger the relationship. Figure 4 illustrates the resulting network of the selected terms.
Page 9 of 16
Analyzing binge watching and marathon watching
Thirty
eigh
th International Conference on Information Systems, Seoul 2017 10
Figure 4. Network Graph
The network graph shows that online television streaming services like Netflix and Hulu relate closely to
binge watching, whereas broadcasting networks like ABC, A&E, AMC and FOX are closer to marathon
watching. HBO, in turn, links to both binge watching and marathon watching (because HBO has also a
video-on-demand service). The parent verb ‘watch’ occurs more likely with marathon watching, whereas
‘view’ appears with both binge watching and marathon watching. Much to our surprise, binge watching
had a strong relationship only with ‘view’. Consistent with the results reported above in Table 2, television
shows such as The walking dead’, ‘Law and Order’ link to marathon watching, whereas ‘House of cards’,
Orange is the new Black’ relate to binge watching.
Television Show Content Analysis
As for the television show attributes, we present the results for both binge watching and marathon
watching in Figure 5. The attributes reported here are rating, starting year, length (no. of
episodes/seasons), continuity, and genre (independent variables). The blue dot on each bar represents a
mean value. Each bar, in turn, illustrates the highest posterior density (HPD) interval, the credible
interval, where the significant estimates exclude zero within the interval. For exact values, please see the
tabular form of the estimates in Appendices A2-A3.
The results suggest that television show ratings favor binge watching over marathon watching. That is,
users are likely to link highly rated television shows to binge watching rather than to marathon watching.
For marathon watching, the relationship is negative and significant only at 75% HPD, indicating that
marathon watching concerns consuming television shows of poorer quality. Concerning the other
television show attributes, ongoing television shows appears to explain marathon watching. This is
and
abc
amc
amr
awk
bng
brk
crm
dxt
dct
fox
gmf
gss
hbo
hsf
hul
lwn
lst
mrt ntf
orn
prt
scn
shr
spr
tnw
tht
thw
twn
viw
wtc
and: A&E
abc: ABC
amc: AMC
amr: American Horror Story
awk: Awkward
bng: Binge
brk: Breaking Bad
crm: Criminal Minds
dxt: Dexter
dct: Doctor Who
fox: FOX
gmf: Game of Thrones
gss: Gossip Girl
hbo: HBO
hsf: House of Cards
hul: Hulu
lwn: Law & Order
lst: Lost
mrt: Marathon
ntf: Netflix
orn: Orange Is the New Black
prt: Pretty Little Liars
scn: Scandal
shr: Sherlock
spr: Supernatural
tnw: Teen Wolf
tht: The Twilight Zone
thw: The Walking Dead
twn: Twenty Four
viw: View
wtc: Watch
Page 10 of 16
Analyzing binge watching and marathon watching
Thirty
eigh
th International Conference on Information Systems, Seoul 2017 11
perhaps because due to lack of user autonomy. The television networks choose what television shows they
air in marathon blocks, and since they typically use them for promotional reasons, they prefer returning
television shows, in which they have invested heavily. The publication year or the number of total
episodes did not explain either binge watching or marathon watching. Interestingly, though, the number
of total seasons correlated positively with marathon watching. As for binge watching, the relationship was
negative, yet insignificant. This indicates that online television streaming services may also include
cancelled television shows or miniseries in their collection. Further, our analysis shows also the difference
in preferences in terms of genre. Of all genres, only drama had a statistically significant positive effect on
binge watching. Comedy, documentary, family, fantasy, game show, and history, in turn, had a negative
effect on binge watching. Our results suggest online television streaming services do not necessarily need
a mixed collection of television shows. For television networks, a diverse range of television shows is
necessary: their users prefer to marathon watch action, horror, romance, thriller, but not comedy.
To sum up, binge watching is a regular behavior that links to high-quality drama, whereas marathon
watching varies irregularly and typically links to returning television shows of various genres.
Marathon model Binge model
Figure 5. Models for Marathon and Binge Tweet Frequencies
Discussion: Implications, Limitations, and Future Research
Following uses and gratifications theory (Katz 1959), we recognize there are users who seek to act
rationally and selectively to adapt their system use to their own purposes. They determine what system
they use, what content they consume, when they consume it, and for how long. Having said this, their
viewing is intentional and goal-directed. However, when seeking to achieve their goals they face system
structures, such as scheduled programming, which limit their user autonomy, that is, their ability to act
on their own reasons, desires, and goals (Friedman 1996). To exemplify the influence of user autonomy on
system use, we focused on a timely usage behavior in the television domain: viewing episodes of a
television show back-to-back. We suggested that users might engage in this behavior by using two
systems: online television streaming services scheduled television. We then mapped these systems onto
binge watching (online television streaming service) and marathon watching (scheduled television), and
explored their outcomes from two perspectives: content selection and time use. The results verify that the
two manifestations differ from one another. This is due to system capabilities, which either facilitate or
undermine user autonomy. As for time use, we found that binge watching fluctuates less than marathon
watching. Online television streaming systems allow their users to fit their viewing around their
commitments. As for content selection, we found that online television streaming services make their
Page 11 of 16
Analyzing binge watching and marathon watching
Thirty
eigh
th International Conference on Information Systems, Seoul 2017 12
users focus on high quality drama, whereas television networks make their users follow returning
television shows of various genres. This suggests that while both online television streaming services and
scheduled television provide their users with the same engagement opportunity, the outcomes the
engagement produces varies between these systems.
The findings have several important implications.
First, there is relatively little research on viewing episodes of the same television show in bulk (Jenner
2016; Pittman and Sheehan 2015; Schweidel and Moe 2016). Few of them acknowledges the behavior has
different manifestations (Glebatis Perks 2014; Jenner 2016; Pittman and Sheehan 2015), yet they have
not discussed the role of the underlying system and its effect on these manifestations. We found the
connection of binge watching and marathon watching to a specific system is trackable. Our findings
illustrate that users who employ scheduled television to engage in marathon blocks, are marathon
watching, whereas those who employ online television streaming services to view episodes of the same
television show in one sitting, engage in binge watching. While this at first sight appears obvious and
irrelevant, distinguishing these manifestations from one another is crucial for examining their outcomes.
Second, our research lend weight to user autonomy. As such, it is, to the best of our knowledge, the first to
attempt to address the differences between binge watching and marathon watching based on user
autonomy. By tapping into user autonomy (Friedman 1996), we explicitly accept that system use is not
completely in users’ own hands. Indeed, the findings imply that scheduled television offer less adaptive
efforts for its users to satisfy their desires. This is because television networks select the television shows
they air in marathon blocks, and based on our findings, these shows have received relatively lower ratings.
Assuming that users prefer viewing high quality television shows it seems that marathon watching is
closer to a benefit-satisficing strategy than a benefit-maximizing strategy (Elie-Dit-Cosaque and Straub
2011). Adopting user autonomy in another IS context is certainly a future research endeavor worth
pursuing. Third, many researchers before us have demanded for improvements in the conceptualization
of system usage. Benbasat and Barki (2007), for example, have proposed researchers should broaden
their horizon to include adaptation, learning, and reinvention behaviors around a system. In line with this
recommendation, Elie-Dit-Cosaque and Straub (2011) have proposed a coping model of user adaptation,
which examines user appraisal strategies and the level of control the users have over the adaptation
events. While we, following Burton-Jones and Straub (2006), focus on distinct system use manifestations,
we acknowledge users’ appraisal with respect to its significance for their wellbeing and the control the
users have over the adopted system and its characteristics to achieve that wellbeing.
All studies have limitations and so does ours. Our sincerest wish, however, is that they will lead to future
research. First, we assess both binge watching and marathon watching as well as their outcomes, but do
not consider general television viewing. Future research could and should simultaneously compare the
outcomes of general television viewing, binge watching, and marathon watching. We would then see how
user autonomy influences all these behaviors. The second limitation is that our Twitter data are sensitive
to holiday seasons, particularly in terms of marathon watching. Ideally, and to avoid this seasonal bias, we
would have liked to extend our data collection period to three years. Third, as correctly pointed by the
reviewers, our analysis would have been even more beneficial, if we examined the tweet sentiments or
grouped users according to their demographics. We acknowledge their feedback and expand our data
collection efforts to make these analyses possible in our forthcoming papers. Finally, we cannot but
recommend that future research takes into account both bingeing watching and marathoning watching.
This is because some television networks deliver television shows via online television streaming services
and scheduled television (e.g., HBO). That is, the television network no longer identifies what system the
users employ for viewing television shows back-to-back.
Page 12 of 16
Analyzing binge watching and marathon watching
Thirty
eigh
th International Conference on Information Systems, Seoul 2017 13
References
Agarwal, R., and Karahanna, E. 2000. "Time Flies when You'Re having Fun: Cognitive Absorption and
Beliefs about Information Technology Usage," MIS Quarterly (24:4), pp. 665-694.
Aslam, S. 2017. "Twitter by the Numbers: Stats, Demographics, and Fun Facts," Omnicore, August 12,
2017, https://www.omnicoreagency.com/twitter-statistics/.
Benbasat, I., and Barki, H. 2007. "Quo Vadis, TAM?" Journal of the Association for Information Systems
(8:4), pp. 211-218.
Burton-Jones, A., and Straub, D. W. 2006. "Reconceptualizing System Usage: An Approach and Empirical
Test," Information Systems Research (17:3), pp. 228-246.
Chen, H., Chiang, R. H. L., and Storey, V. C. 2012. "Business Intelligence and Analytics: From Big Data to
Big Impact," MIS Quarterly (36:4), pp. 1165-1188.
Chen, M., and Chao, Q. 1999. "Monte Carlo Estimation of Bayesian Credible and HPD Intervals," Journal
of Computational and Graphical Statistics (8:1), pp. 69-92.
Conlin, L. T. 2015. "There Goes the Weekend: Understanding Television Binge-Watching," Dissertation,
The University of Alabama.
Conlin, L. T., Billings, A. C., and Averset, L. 2016. "Time Shifting Vs. Appointment Viewing: The Role of
Fear of Missing Out within TV Consumption Behaviors," Communication and Society (29:4), pp.
151-164.
Csikszentmihalyi, M. 1991. Flow, the Psychology of Optimal Experience, New York: HarperPerennial.
De Feijter, D., Vassilis-Javed, K., and van Gisbergen, M. 2016. "Confessions of a 'Guilty' Couch Potato:
Understanding and using Context to Optimize Binge Watching Behavior," Proceedings of the ACM
International Conference on Interactive Experiences for TV and Online Video, June 22-24, 2016,
pp. 59-67.
Devasagayam, R. 2014. "Media Bingeing: A Qualitative Study of Psychological Influences," Proceedings of
the Marketing Management Association, March, pp. 40.
Elie-Dit-Cosaque, C. M., and Straub, D. W. 2011. "Opening the Black Box of System Usage: User
Adaptation to Disruptive IT," European Journal of Information Systems (20:5), pp. 589-607.
Friedman, B. 1996. "Value-Sensitive Design," Interactions (3:6), pp. 16-23.
Gelman, A., Carlin, J., Stern, H., Dunson, D., and Rubin, D. 2014. Bayesian Data Analysis, Boca Raton:
Chapman & Hall / CRC.
Glebatis Perks, L. 2014. Media Marathoning: Immersions in Morality, New York: Lexington Books.
Goodhardt, G. J., Ehrenberg, A. S., and Collins, M. A. 1987. The Television Audience: Patterns of
Viewing, Gower: Aldershot.
Graves, M. 2015. ""Chalk One Up for the Internet: It has Killed Arrested Development": The Series'
Revival, Binge Watching and Series/Fan Critic and Antagonism," in A State of Arrested
Development: Critical Essays on the Innovative Television Comedy, Kristin M. Barton (ed.),
Jefferson: McFarland & Company Inc., pp. 224-236.
Green, P. E., and Srinivasan, V. 1978. "Conjoint Analysis in Consumer Research: Issues and Outlook,"
Journal of Consumer Research (5:2), pp. 103-123.
Harris Interactive. 2013. "Americans Taking Advantage of Ability to Watch TV on their Own Schedule,"
April 8, 2013, http://www.prnewswire.com/news-releases/americans-taking-advantage-of-ability-
to-watch-tv-on-their-own-schedules-201924871.html.
Hewitt, M. 2015. "The Watcher: On Thanksgiving, it's Turkey with a Side of TV Marathons," The Orange
County Register, November 23, 2015, http://www.ocregister.com/2015/11/23/the-watcher-on-
thanksgiving-its-turkey-with-a-side-of-tv-marathons/.
Jenner, M. 2016. ""Is this TVIV? on Netflix TVIII and Binge-Watching"," New Media and Society (18:2),
pp. 257-273.
Jenner, M. 2015. "Binge-Watching: Video on Demand, Quality TV and Mainstreaming Fandom,"
International Journal of Cultural Studies (Online first), pp. 1-17.
Johnson, D. H. 1999. "The Insignificance of Statistical Significance Testing," The Journal of Wildlife
Management (63:3), pp. 763-772.
Katz, E. 1959. "Mass Communications Research and the Study of Popular Culture: An Editorial Note on a
Possible Future for this Journal," Studies in Public Communication (2), pp. 1-6.
Page 13 of 16
Analyzing binge watching and marathon watching
Thirty
eigh
th International Conference on Information Systems, Seoul 2017 14
Katz, E., Blumler, J. G., and Gurevitch, M. 1974. "Utilization of Mass Communication by the Individidual,"
in The Uses of Mass Communications: Current Gratifications Research, Jay G. Blumler and Elihu
Katz (eds.), Beverly Hills: Sage Publications, pp. 19-32.
Kompare, D. 2002. "I'Ve seen this One before: The Cosntrusction of 'Classic TV' on Cable Television," in
Small Screens, Big Ideas: Television in the 1950s, Janet Thumin (ed.), New York: I.B. Tauris
Publishers, pp. 19-34.
Matrix, S. 2014. "The Netflix Effect: Teens, Binge Watching, and on-Demand Digital Media Trends,"
Jeunesse: Young People, Texts, Cultures (6:1), pp. 119-138.
Morstatter, F., Pfeffer, J., Liu, H., and Carley, K. M. 2013. "Is the Sample Good enough? Comparing Data
from Twitter's Streaming API with Twitter's Firehose," Proceedings of the Seventh International
AAAI Conference on Weblogs and Social Media, 8-11 July, 2013, pp. 400-408.
Netflix. 2013. "Netflix Declares Binge Watching is the New Normal," PR Newswire, December 13, 2013,
http://www.prnewswire.com/news-releases/netflix-declares-binge-watching-is-the-new-normal-
235713431.html.
Nielsen. 2013. ""Binging" is the New Viewing for Over-the-Top Streamers," September 18, 2013,
http://www.nielsen.com/us/en/insights/news/2013/binging-is-the-new-viewing-for-over-the-top-
streamers.html.
O'Hara, R. B., and Kotze, D. J. 2010. "Do Not Log-Transform Count Data," Methods in Ecology and
Evolution (1:2), pp. 118-122.
Pingree, S., Hawkins, R. P., Bush Hitchon, J. C., Gilligan, E., Radler, B., Kahlor, L., Gorham, B., Kolbeins,
G. H., Schmidt, T., and Kannaovakun, P. 2001. "If College Students are Appointment Television
Viewers..." Journal of Broadcasting & Electronic Media (45:3), pp. 446-463.
Pittman, M., and Sheehan, K. 2015. "Sprinting a Media Marathon: Uses and Gratifications of Binge-
Watching Television through Netflix," First Monday (20:10), October 5.
Pullar-Strecker, T. 2015. "Evening Broadband Hit by 'Netflix Effect'," Stuff, April 24, 2015,
http://www.stuff.co.nz/business/industries/68027358/Evening-broadband-hit-by-Netflix-effect.
Ruggiero, T. E. 2000. "Uses and Gratifications Theory in the 21st Century," Mass Communication &
Society (3:1), pp. 3-37.
Rust, R. T., and Alpert, M. I. 1984. "An Audience Flow Model of Television Viewing Choice," Marketing
Science (3:2), pp. 113-124.
Ryan, R. M., and Deci, E. L. 2006. "Self-Regulation and the Problem of Human Autonomy: Does
Psychology Need Choice, Self-Determination, and Will?" Journal of Personality (74:6), pp. 1557-
1585.
Sanchez, G. 2013. Handling and Processing Strings in R, Berkeley: Trowchez editions.
Schweidel, D. A., and Moe, W. W. 2016. "Binge Watching and Advertising," Journal of Marketing (80:5),
pp. 1-19.
Statista. 2017. "Number of Monthly Active Twitter Users Worldwide from 1st Quarter 2010 to 4th Quarter
2016," Statista, https://www.statista.com/statistics/282087/number-of-monthly-active-twitter-
users/.
Vander Verff, T. 2016. "3 Reasons Why Netflix can't Keep its Ratings a Secret Forever," Vox, January 22,
2016, https://www.vox.com/2016/1/22/10813288/netflix-ratings-revealed-nbc.
Wallenstein, A. 2014. "'House of Cards' Binge-Watching: 2% of U.S. Subs Finished Entire Series Over
First Weekend," Variety, February 20, 2014, http://variety.com/2014/digital/news/house-of-cards-
binge-watching-2-of-u-s-subs-finished-entire-series-over-first-weekend-1201114030/.
Webster, J. G. 2014. The Marketplace of Attention: How Audiences Take Shape in a Digital Age?,
Cambridge: MIT Press.
Webster, J. G., and Newton, G. D. 1988. "Structural Determinants of the Television News Audience,"
Journal of Broadcasting & Electronic Media (32:4), pp. 381-389.
Wortham, J. 2011. "Feel Like a Wallflower? Maybe it's Your Facebook Wall," The New York Times, April
9, 2011, http://www.nytimes.com/2011/04/10/business/10ping.html.
Page 14 of 16
Analyzing binge watching and marathon watching
Thirty
eigh
th International Conference on Information Systems, Seoul 2017 15
Appendices
Table A1. Data Collection Search Result
Keyword No. of Tweets Keyword No. of Tweets
binge 257251 #bingewatching 4504
marathon watch 116728 binge viewing 2689
marathon watching 115189 marathon viewing 2360
TV marathon 63329 broadcast marathon 1567
binge watching 61601 binge view 506
binge watch 26016 #bingeviewing 298
bingeing 9787
Total Number of Tweets 661825
Table A2. Estimates for the Marathon Model
6
Variable
mean
sd
2.5%
25%
50%
75%
97.5%
(Intercept)
5.047
0.518
0.538
0.547
5.041
0.585
0.624
IMDb
Rating
-
0.146
0.093
-
0.330
-
0.208
-
0.146
-
0.084
0.037
Starting
Year
0.017
0.010
-
0.002
0.010
0.017
0.023
0.036
Number of
Seasons
0.098
0.028
0.043
0.079
0.098
0.081
0.106
Number of
Episodes
0.000
0.001
-
0.001
-
0.001
0.000
0.000
0.001
Action
0.492
0.192
0.127
0.360
0.488
0.619
0.231
Adventure
0.238
0.242
-
0.331
0.073
0.234
0.398
1.040
Animation
-
0.433
0.227
-
1.054
-
0.655
-
0.443
-
0.222
0.162
Comedy
-
0.014
0.154
-
0.453
-
0.163
-
0.012
0.096
0.420
Crime
0.150
0.197
-
0.332
0.025
0.149
0.281
0.054
Documentary
-
0.767
0.458
-
1.949
-
1.223
-
0.805
-
0.356
0.433
Drama
0.190
0.126
-
0.167
0.068
0.131
0.217
0.550
Family
0.288
0.201
-
0.128
0.151
0.283
0.419
1.001
Fantasy
-
0.067
0.240
-
0.730
-
0.300
-
0.073
0.112
0.439
Game
Show
-
0.659
0.044
-
1.823
-
1.095
-
0.681
-
0.251
0.441
History
0.114
0.503
-
1.185
-
0.332
0.094
0.439
0.502
Horror
0.634
0.283
0.099
0.440
0.626
0.169
0.555
Music
0.130
0.520
-
0.827
-
0.229
0.074
0.323
0.185
Mystery
0.357
0.218
-
0.096
0.209
0.510
0.502
0.142
Reality
-
0.037
0.360
-
0.724
-
0.280
-
0.042
0.140
0.475
Romance
0.590
0.169
0.267
0.476
0.587
1.012
0.274
SciFi
-
0.845
0.456
-
2.043
-
1.297
-
0.877
-
0.427
0.362
Sport
-
0.011
0.437
-
1.121
-
0.450
-
0.053
0.266
0.281
Thriller
0.153
0.274
0.283
0.622
2.094
0.335
0.706
War
0.406
0.180
-
1.641
-
0.231
0.367
0.286
3.010
Ongoing
0.672
0.134
0.406
0.583
0.672
1.096
0.281
Reciprocal
dispersion
1.078
0.074
0.610
1.005
1.075
0.144
0.248
Mean
PPD
539.152
206.923
339.071
426.028
492.981
590.851
1023.355
L
og
-
posterior
-
1230
1
-
1239
-
1232
-
1230
-
1227
-
1223
6
For the significant estimates, zero lies outside 2.5% -97.5% interval.
Page 15 of 16
Analyzing binge watching and marathon watching
Thirty
eigh
th International Conference on Information Systems, Seoul 2017 16
Table A
3
.
Estimates for
the Binge M
odel
7
Variable
mean
sd
2.5%
25%
50%
75%
97.5%
(Intercept)
0.390
0.546
-
0.018
0.067
0.386
2.022
3.063
IMDb
Rating
0.322
0.091
0.198
0.280
0.322
0.365
0.044
Starting
Year
-
0.023
0.010
-
0.043
-
0.030
-
0.023
-
0.016
-
0.002
Number of
Seasons
0.039
0.029
-
0.017
0.020
0.039
0.058
0.096
Number of
Episodes
0.000
0.001
-
0.001
0.000
0.000
0.001
0.002
Action
-
0.400
0.204
-
0.964
-
0.600
-
0.406
-
0.206
0.135
Adventure
0.398
0.380
-
0.160
0.218
0.395
0.572
0.272
Animation
-
0.535
0.222
-
1.141
-
0.752
-
0.544
-
0.326
0.077
Comedy
-
0.586
0.171
-
1.069
-
0.752
-
0.586
-
0.424
-
0.099
Crime
-
0.155
0.176
-
0.646
-
0.327
-
0.159
0.013
0.350
Documentary
-
2.330
0.579
-
3.832
-
2.913
-
2.371
-
1.803
-
0.578
Drama
0.349
0.141
0.074
0.254
0.349
0.444
0.628
Family
-
1.396
0.217
-
1.987
-
1.609
-
1.401
-
1.192
-
0.767
Fantasy
-
0.793
0.248
-
1.479
-
1.037
-
0.797
-
0.557
-
0.076
Game
Show
-
1.691
0.421
-
2.835
-
2.108
-
1.707
-
1.293
-
0.450
History
-
2.436
0.048
-
3.667
-
2.914
-
2.475
-
1.999
-
0.977
Horror
-
0.166
0.410
-
0.934
-
0.451
-
0.178
0.072
0.469
Music
0.408
0.480
-
0.300
0.178
0.397
0.624
0.442
Mystery
-
0.335
0.211
-
0.919
-
0.543
-
0.338
-
0.131
0.184
Reality
-
0.795
0.312
-
1.667
-
1.100
-
0.799
-
0.493
0.090
Romance
0.148
0.175
-
0.272
0.042
0.145
0.263
0.499
SciFi
-
0.298
0.403
-
1.345
-
0.701
-
0.329
0.070
0.650
Sport
-
0.181
0.444
-
1.296
-
0.625
-
0.227
0.144
0.192
Thriller
0.353
0.248
-
0.172
0.185
0.050
0.516
0.203
War
0.083
0.202
-
2.153
-
0.716
0.059
0.616
0.597
Ongoing
0.143
0.135
-
0.179
0.075
0.144
0.235
0.404
Reciprocal
dispersion
1.045
0.074
0.589
0.674
1.041
0.121
0.225
Mean
PPD
214.292
36.842
158.519
190.105
210.218
233.211
293.033
L
og
-
posterior
-
1079
1
-
1087
-
1081
-
1078
-
1076
-
1072
7
For the significant estimates, zero lies outside 2.5% -97.5% interval.
Page 16 of 16
  • ... Some publications reviewed in this article have linked BW to professionally produced quality television shows available on subscription services (Hargraves 2015;Jenner 2017). Of these television shows, dramas have received the most attention (Upreti et al. 2017), and consuming these hour-long episodes are likely to explain why BW seems inherently excessive. Yet, we claim that this is not necessarily the case as BW goes beyond popular content adhering to traditional molds of television (Volpe 2017). ...
    Article
    Binge watching serialized video content is a phenomenon that has triggered interest from diverse research fields. Despite the progress researchers have made across different areas, a grounded conceptualization and definition of binge watching is still lacking. In this article, we conduct a transdisciplinary literature review to identify continuity and viewer autonomy as the two fundamental attributes underlying binge watching. Then, using these attributes as conceptual anchors, we offer a convergent definition and categorize the existing binge-watching definitions in the literature. The results of this categorization reveal that the vast majority of the definitions used in the literature fail to distinguish binge watching from other viewing practices such as casual viewing, single-episode appointments, and marathon appointments. We discuss the implications and, to move the binge-watching research forward, conclude with recommendations and an agenda for future research.
  • Thesis
    Full-text available
    The advent of digitalization has brought a massive proliferation of unstructured data, producing vast repositories of textual data, from various sources, such as Web sites, academic publications, news articles, blog posts, e-mail, corporate communication platforms, reports, and social media feeds. This proliferation coupled with the upsurge in mobile and Web technologies alongside ever- improving connectivity has led to various digital platforms and applications rapidly achieving mass-market penetration. With the production of textual and other forms of unstructured data certain to continue at unprecedented rates for the foreseeable future, this availability on massive scale presents both opportunities and challenges that researchers and practitioners must address. Ability to utilize text data on a large scale not only provides better coverage in terms of sample size but also opens opportunities to build a deeper understanding of phenomena that otherwise are simply unobservable, "hidden in the noise.'' However, as the world races towards high-volume production, distribution, and consumption of digital text, information systems (IS) researchers are proving slow to start reaping the potential of analyzing textual data. There is an urgent need for methods and techniques that can meet the challenge of analyzing vast bodies of textual data. In an effort to demonstrate potential application of text-mining methods in information systems research, the dissertation presents essays that address large-scale text-based datasets' use in literature analysis and studies of system-specific behavioral outcomes. The first essay deals with identifying the research themes presented in a large body of publications on cloud computing, and the second essay demonstrates the machine-based classification of papers in leading information- systems journals. Of the behavior-focused pieces, the third essay utilizes user-generated content to illustrate system-driven viewing outcomes in the context of binge watching of television shows, and the final essay examines a large volume of content connected with a business-to-business Web portal, reporting on a study of browsing-device-linked differences in interest in marketing material. In addition to the individual essays, the dissertation contributes to the scholarly discussion of text- mining research issues in three important ways. Firstly, it presents a conceptual framework that aids in revealing the fundamentals of text-mining research in terms of two dimensions: research objective and level of text analysis. Secondly, the four essays provide concrete demonstrations of various suitable applications of text-mining. Finally, the dissertation examines the implications of the work, highlighting specific issues and challenges pertaining to text-mining research. The findings and implications of this work should benefit IS researchers and practitioners striving to exploit large volume of textual data.
  • Article
    Full-text available
    The current study employed a national sample in order to investigate the phenomenon of fear-of-missing-out (FoMO), the apprehension associated with the fear that other people are having a pleasurable experience that one is not a part of. The current study investigated the role that FoMO plays in TV viewing habits, particularly binge-watching and the consumption of one-time megaevents. Results indicated that FoMO predicts the pace at which people choose to watch TV, social media use as it relates to TV, and whether they are likely to watch some one-time TV programs—such as sporting events like the Super Bowl.
  • Conference Paper
    Viewers more frequently watch television content whenever they want, using devices they prefer, which stimulated "Binge-watching" (consecutive viewing of television programs). Although binge-watching and health concerns have been studied before, the context in which binge-watching takes place and possibilities to use context to optimize binge-watching behavior have not. An in-situ, smartphone monitoring survey among Dutch binge-watchers was used to reveal context factors related to binge-watching and wellbeing. Results indicate that binge-watching is a solitary activity that occurs in an online socially active context. The amount of time spent binge-watching (number of episodes) correlates with the amount of free time and plays an important role in the effect of binge-watching on emotional wellbeing. Considering the difficulty viewers have to create an optimal viewing experience, these context factors are used as a framework to be able to design and promote a recommendation tool for TV streaming services to create a more optimal binge-watching experience.
  • Article
    How users consume media has shifted dramatically as viewers migrate from traditional broadcast channels toward online channels. Rather than following the schedule dictated by television networks and consuming one episode of a series each week, many viewers now engage in binge watching, which involves consuming several episodes of the same series in a condensed period of time. In this research, the authors decompose users' viewing behavior into (1) whether the user continues the viewing session after each episode viewed, (2) whether the next episode viewed is from the same or a different series, and (3) the time elapsed between sessions. Applying this modeling framework to data provided by Hulu.com, a popular online provider of broadcast and cable television shows, the authors examine the drivers of binge watching behavior, distinguishing between user-level traits and states determined by previously viewed content. The authors simultaneously investigate users' response to advertisements. Many online video providers support their services with advertising revenue; thus, understanding how users respond to advertisements and how advertising affects subsequent viewing is of paramount importance to both advertisers and online video providers. The results of the study reveal that advertising responsiveness differs between bingers and nonbingers and that it changes over the course of online viewing sessions. The authors discuss the implications of their results for advertisers and online video platforms.
  • Article
    Business intelligence and analytics (BI&A) has emerged as an important area of study for both practitioners and researchers, reflecting the magnitude and impact of data-related problems to be solved in contemporary business organizations. This introduction to the MIS Quarterly Special Issue on Business Intelligence Research first provides a framework that identifies the evolution, applications, and emerging research areas of BI&A. BI&A 1.0, BI&A 2.0, and BI&A 3.0 are defined and described in terms of their key characteristics and capabilities. Current research in BI&A is analyzed and challenges and opportunities associated with BI&A research and education are identified. We also report a bibliometric study of critical BI&A publications, researchers, and research topics based on more than a decade of related academic and industry publications. Finally, the six articles that comprise this special issue are introduced and characterized in terms of the proposed BI&A research framework.
  • Article
    Full-text available
    "Binge-watching" represents a radical shift for twenty-first century media consumption. Why do people select this method of television viewing? A survey administered to 262 television binge-watchers identified factors that influence binge watching, several of which are somewhat different than factors impacting other types of television viewing. Factors salient for regular bingers are relaxation, engagement, and hedonism. For those who plan ahead to binge, program quality (aesthetics) and the communal aspect (social) also come into play. Those who binge on an entire series in one or two days value engagement, relaxation, hedonism, and aesthetics. We also discuss the theoretical implications and future development of uses and gratifications.
  • Article
    Full-text available
    This article explores the concept of the binge as viewing protocol associated with fan practices, industry practice and linked to ‘cult’ and ‘quality’ serialised content. Viewing binge-watching as an intersection of discourses of industry, audience and text, the concept is analysed here as shaped by a range of issues that dominate the contemporary media landscape. In this, factors like technological developments, fan discourses and practices being adopted as ‘mainstream’ media practice, changes in the discursive construction of ‘television’ and an emerging video-on-demand industry contribute to the construction of binge-watching as deliberate, self-scheduled alternative to ‘watching TV’.