"I Can Watch What I Want": A Diary Study of On-Demand

and Cross-Device Viewing

Jacob M. Rigby1 , Duncan P. Brumby1 , Sandy J.J. Gould2 , Anna L. Cox1

1
UCL Interaction Centre, University College London, London, WC1E 6EA, UK
2 School of Computer Science, University of Birmingham, Birmingham, B15 2TT, UK
1 {j.rigby.14, d.brumby, anna.cox}@ucl.ac.uk, 2 [email protected]

ABSTRACT are useful for giving a general impression of the popularity

In recent years, on-demand video services, such as Netflix and of on-demand services. However, such surveys can lack the
Amazon Video, have become extremely popular. To under- necessary level of granularity to unpick what is driving these
stand how people use these services, we recruited 20 people viewing practices. For example, Ofcom [12] suggest that the
from nine households to keep a viewing diary for 14 days. To TV is still the most popular way to view, but that 21% of
better understand these household viewing diaries, in-depth the online population choose to watch on a phone, 23% on a
interviews were conducted. We found that people took advan- tablet and 33% on a computer at least once a month. These
tage of the freedom and choice that on-demand services offer, surveys provide useful data but do not ask about important
watching on different devices and in different locations, both contextual and situational factors that might be affecting why
in the home and outside. People often watched alone so they people choose to watch on one device over another. Does this
could watch what they wanted, rather than coming together to decision of which device to use depend on where the person
watch something of mutual interest. Despite this flexibility, the is watching? Who they are watching with? What time of day
evening prime time continued to be the most popular time for they are watching? What they are watching? We focus here on
people to watch on-demand content. Sometimes they watched what motivates such decisions that people take when viewing.
for extended periods, and during interviews concerns were
In this paper, we describe the results of a diary study con-
expressed about how on-demand services make it far too easy
ducted to provide a detailed snapshot of everyday viewing
to watch too much and that this is often undesirable.
behaviours using on-demand services. The work presented
CCS Concepts
here extends an initial analysis of these viewing diaries [16]
by describing the prevalence of different practices surround-
•Human-centered computing → Empirical studies in HCI;
ing on-demand viewing. In particular, we focus our analysis
Human computer interaction (HCI);
on unpicking differences in viewing behaviour on handheld
Author Keywords
mobile devices and non-mobile devices. In this paper we also
present the results of in-depth interviews that were conducted
On-demand video; film; television; video; streaming; diary
to better understand these household viewing diaries. These
study; mobile viewing; IPTV; VoD; SVoD; binge watching
interviews focused on understanding what motivated different
INTRODUCTION
viewing behaviours: why people choose to view on particular
devices, watch in different locations, and watch alone or to-
Consuming video through on-demand video services has be-
gether. We also develop an understanding of people’s positive
come a popular activity in recent years. According to the
and negative perceptions of on-demand services.
Nielsen company, 43% of people globally watch some kind
of on-demand video at least once a day [3]. Subscriptions to RELATED WORK
paid services (e.g., Netflix and Amazon Video) are rising year Prior to the advent of on-demand video services, viewers had
on year, and total on-demand viewing as a percentage of all limited choice about what they watched, and when and where
viewing (including viewer-recorded content) is also increasing they watched it. Previous research from this era gives us
annually in the UK [12]. an insight into "traditional" linear TV viewing practices. For
With the rise in popularity of on-demand services, what impact instance, evening viewing after the working day was especially
is this having on how people consume video content? Large- popular, particularly in the living room [18]; people watched
scale surveys, such as those from Ofcom [12] and Nielsen [3], TV regularly, often for multiple hours per day [8]; and personal
viewing schedules were based around broadcast schedules,
which in turn influenced other household activities [6].
This work is licensed under a Creative Commons
Considering the current popularity of on-demand video ser-
Attribution International 4.0 License.
vices and mobile viewing, surprisingly little HCI literature has
addressed it. An early study by O’Hara et al. [13] sought to
TVX ’18 June 26–28, 2018, SEOUL, Republic of Korea better understand how watching video on mobile devices fits
© 2018 Copyright held by the owner/author(s). into people’s lives. In agreement with similar findings by Of-
ACM ISBN 978-1-4503-5115-7/18/06. com [11], O’Hara et al. [13] found that portability and fitting
DOI: https://doi.org/10.1145/3210825.3210832

69
in with other peoples’ schedules were important factors. They concerned with the technicalities of delivering video to con-
also found other reasons for watching, such as simply passing sumers, it does offer some high-level insights into viewer
time, and being able to be present with others while still con- behaviour. Nogueira et al. found that users interacted with this
suming video privately. Conversely, O’Hara et al. [13] also service throughout the day, though viewing was most popular
found that mobile video was used to disengage with others and in the evening. They also found that users exhibited a large
signifying the wish to be left alone. This study mainly focused amount of "zapping" behaviour in order to select content, simi-
on the motivations rather than establishing prevalence, perhaps lar to "channel surfing", taking on average 2.5 minutes to settle
due to being conducted back in 2007, before powerful mobile on something. However, their data does not offer insights into
devices and on-demand video services were commonplace. mobile viewing, and only covers use of a single video service.
McNally and Harrington [9] conducted a more recent study In summary, previous research has provided useful perspec-
on how teens and millennials consume mobile video, again fo- tives into how people use on-demand video services. A com-
cusing on motivations rather than prevalence. They found that mon theme is that people value and and take advantage of
motivations depended on mood and emotional state. McNally freedom from the broadcast schedule, allowing them to choose
and Harrington also investigated how content was chosen, find- viewing times that suit them. Furthermore, much of the lit-
ing that it was based on the level of stimulation provided, as erature reveals a strong social element to watching TV. Be it
well as video length and amount of engagement required. watching together, selecting content together, or discussing
shows with friends and colleagues, social factors appear to
Bury and Li [2] conducted a survey study in 2013 into different
affect viewing practices. Prior research also gives us a limited
ways of consuming TV. They found that mobile viewing was
perspective on viewing on mobile devices, specifically regard-
unpopular, with 70% of respondents never having used mobile
ing motivations for doing so, which are many and varied.
devices for viewing. Those that participated in mobile viewing
mainly did so when travelling and commuting. However, this While the phenomena of on-demand viewing and mobile view-
seems to have changed in recent years, with mobile viewing ing are strongly coupled, they have not been investigated to-
growing in popularity [12]. This study also clearly shows a gether from an HCI perspective using recent, real-world data,
general shift away from live TV viewing to online viewing. which would allow us to develop deeper behavioural insights.
Furthermore, we do not know exactly how people are using
Barkhuus and Brown [1] conducted in-depth interviews to
these services throughout the day over longer periods of time,
understand how TV watching was changing as a result of new across different devices and services, and what motivates par-
technologies. In particular, they focused on personal video ticular viewing behaviours. In the following sections, we
recorders (PVR) and internet downloads, as this study was present the results of a diary study with interviews that was
conducted in 2009, before on-demand video services were conducted over a 14-day period with 20 people from nine
common. They found that most participants who used a PVR households. We asked people to record the details of each
system had moved away from watching live TV almost en- time someone viewed on-demand content in the household.
tirely, preferring to queue up recordings from their downloaded These diaries focused on when and where viewing took place,
library. This freedom from the TV schedule was particularly
as well as which services and devices were being used. Pre-
valued by those with non-standard work schedules.
and post-study interviews were also conducted to further probe
Irani et al. [7] conducted a diary study of people’s viewing and understand these present-day viewing practices.
habits. This study examined the temporality of viewing in 14
households, which included the use of time shift and early METHOD
on-demand services. They found that viewing was typically Participants
based around the rhythms of individuals’ lives, households, Ten UK households who watched at least five hours of on-
and peers. The ability to choose when to watch could help demand content a week were recruited through word of mouth
align televisual schedules, allowing members of a household to and advertisements (see Table 1 for breakdown). One (house-
watch together. There was also much discussion in households hold C) withdrew, leaving 20 remaining participants from nine
about what to watch and about the content of a show. Irani et households. Mean age was 29.8 (SD = 13.8). Households were
al. also found that TV content was used as a background to paid £100 (~$137) for 14 days of continuous participation.
other tasks, and to fill gaps of unscheduled time.
Materials
A study by Vanattenhoven and Geerts [20] also looked at how Households chose either a paper or digital diary. Seven chose
different ways of consuming media occurred around the house digital and two chose paper. For the digital diary, data was en-
via qualitative interviews, including on-demand content. They tered into an online form using any device with a web browser.
noted that viewing depended on the context of other things Results were stored in a spreadsheet. For the paper diaries,
happening in the household. They found that on-demand custom diary booklets were created for each household. After
viewing typically involved "heavier" content requiring more data collection, they were digitised in the same format as the
focus (e.g., films and TV series), and took place in the evening. digital ones for ease of analysis. Diaries were designed to
In contrast, broadcast TV typically involved "lighter" content make data entry as easy as possible, e.g. with checkboxes for
(e.g., news), which was watched while doing other tasks. names, locations, and services.
Nogueira et al. [10] analysed a large dataset from a Portuguese Participants completed information about each viewing ses-
IPTV operator. While the insights from this work are largely sion, defined as a period of viewing with at least 30 minutes of

70
 y
Re old

nt
Lo ses

lit
n

pa

na
h

tio
on

r
de
ci
se

es

io
sp

ca

rti
ou

ge

en
ot

at
Pa
H

N
A 27 Birmingham Cohabiting A1 57 F British
couple A2 68 M British
B 36 Birmingham Parents and B1 33 M British
their children B2 38 F British
B3 8 M British
B4 4 F British
B5 2 M British
C - - Withdrew - - - - Figure 1: Distribution of viewing screens
D 22 London Cohabiting D1 32 M Spanish
couple D2 29 F Spanish
E 18 London Cohabiting E1 31 M Danish
couple E2 29 F Danish
F 24 London Cohabiting F1 27 F Mexican RESULTS
(others not Participants created 202 diary entries in total. Of these, 24 said
participating) that no on-demand service usage occurred that day, leaving
G 14 London Cohabiting G1 32 M Italian 178 remaining entries describing on-demand viewing. Mean
couple G2 32 F Italian entries per household was 20.6 (SD = 9.1). These diaries
H 15 Oxford Cohabiting H1 27 F British captured 188:36:00 (HH:MM:SS) of viewing time, with a
friends H2 30 F British mean of 20:57:20 per household (SD = 08:10:11). We provide
I 7 London Cohabiting I1 27 F German both a quantitative and qualitative analysis of diary entries. For
couple I2 35 M British all statistical analysis we opted to aggregate data by household
J 15 London Cohabiting J1 31 M German
rather than at the level of the individual participant. Interviews
couple J2 33 M British were also conducted with participants at the end of the study
to learn more about their diary entries and on-demand viewing
Table 1: Participant household profiles habits. These interviews were transcribed and were analysed
thematically using an inductive coding approach.
In this section, we present data from both the diary entries and
non-viewing activity either side to allow for short to medium
direct quotations from the thematic analysis of our interview
breaks. Participants were required to fill in basic information
data. We cluster this around eight different headings. First,
about their viewing: who was present, start and finish times,
viewing screens, is where we consider which devices people
what was watched, how long for, devices and services used,
chose for viewing and why. Second, viewing location, where
location, and breaks they took. They were also asked to justify
we consider the places both inside and outside of the home
and explain their responses, where appropriate.
where people chose to view. Third, viewing time of day and du-
For this study, on-demand content is defined as that which is ration, where we consider how viewing fits into people’s daily
accessed at the viewer’s convenience. This includes catch-up activities and how long they view for. Fourth, services used,
services (e.g. BBC iPlayer), subscription services (e.g. Netflix, where we consider exactly which on-demand services people
Amazon), short-form content (e.g. Youtube, Facebook), and used to access content. Fifth, watching alone and watching
content downloaded or recorded onto computers or personal together, where we consider participants’ co-viewing habits.
video recorders (e.g. TiVo). We also focus on two groups of Sixth and seventh, positive perceptions of on-demand viewing
devices: non-mobile devices (TV, desktop computer, laptop and negative perceptions of on-demand viewing, where we
computer) and handheld mobile devices (phone and tablet). explore what people like and dislike about these platforms.
Finally, binge watching, where we focus on how on-demand
Procedure services can facilitate viewing a lot of content in one session,
After recruiting participants, a preliminary interview was con- and how people think about and define binge watching.
ducted to ascertain their general on-demand viewing habits
and motivations. They were then briefed on how to enter data Viewing Screens
in their diaries. Participants were requested to create at least We first focus on the kind of screen that participants used
one diary entry per day, but this could simply be to say that no to view content. Diary entries fell into five distinct viewing
viewing took place. For each household, one participant was device categories, pre-specified in the participants’ diaries.
nominated to be responsible for the diary and complete it on These are shown in Figure 1. We further collapsed these
behalf of others if necessary, though other household members device categories into two distinct groups: non-mobile devices
were encouraged to fill in the diary as well. During the study, (TV, desktop computer, laptop computer) and handheld mobile
participants were sent SMS reminders every evening to encour- devices (phone and tablet). Of the 178 entries, 55 (29.9%)
age participation. After the study was over another interview contained viewing on a handheld mobile device (i.e., phone
was conducted to ask them about their experiences with using or tablet). Households reported more viewing sessions on
the diary, as well as to explain particular behaviours. non-mobile devices (M = 14.0, SD = 8.3) than on handheld

71
mobile devices (M = 5.8, SD = 6.7). However, this difference B1: I was talking to [my friend] about this earlier and he
was not significant, t(8) = 1.98, p = 0.08. said every night he’ll sit and watch a film on his phone.
He’ll sit there, like, next to [his wife] and she’ll sit there
To further understand how people chose a viewing device, we watching something he’s not interested in and he’ll sit
next look to diary their entries and what was said during the there and watch a film or watch videos on YouTube, some-
end of study interviews. It became clear that different viewing thing to do with work, whatever. And his phone is the
devices were chosen for different reasons. For example, partic- same size as mine. I couldn’t imagine watching a whole
ipant A1 described how she and A2 (her partner) would choose film, just because it’s too small.
their tablets when they wanted to watch content individually,
while still being together in the same room. When asked further about mobile viewing, he clarified:
A1: [We watch] the stuff on the tablet singly — we both B1: I don’t really get much pleasure holding the tablet to
watch different things on that — but on the TV we tend to watch something. [...] It doesn’t interest me, I’d rather sit
put something on that we both want to watch. and watch it on the telly or not bother. [...] One, you’ve
got to hold it and two, the size of the screen.
This was later clarified:
However, he did see a benefit to mobile viewing in keeping
A1: I can watch what I want to watch. We both put our children occupied:
earphones on and we can then watch our own watch-
ing[...]. The TV, that’s our bit of relaxation together. But B1: What I would say about the tablet and the phones,
our little bit of YouTube is what we do on our own. though, is having the kids, when you’re out and, say
you’re going for a meal or something like that, having
Portability was another factor. Participant B1, an eight-year- the phone or tablet with video or like you say, YouTube,
old child, said he liked to be able to watch anywhere, instantly: is really quite handy because it does keep them occupied.
Interviewer: Why do like to watch it on a tablet?
While the laptop was the most popular viewing screen, par-
B3: Because I can take it anywhere. TV, [...] you have to ticipants consistently said that they would prefer to watch on
leave it there. And [other devices] take loads of time to a television. One of the main reasons for this was the bigger
set up if you take it somewhere. screen, but participants also liked the associated comfortable
Participant J1 said that the device could dictate the content that seating. Household A said how watching on the TV was just
was viewed, with phone viewing only being for short clips: part of their routine:

J1: Most of the time the phone is usually for only shorter Interviewer: Why do you watch, for instance, Better Call
snippets it’s like YouTube, or Twitter things... like really Saul on the TV?
short up, to five minutes or so. [...] if I’m taking the time A2: Bigger screen.
to watching something for longer, I can also take the time A1: Bigger screen, yeah. [...] And it’s our sort of evening
to just sit on the couch and relax. routine, we come in [the living room], we sit down and
Participant F1 said the phone was her preferred device in many we watch TV and that’s... yeah, it’s our routine really.
cases, also due to the immediacy of it: Participant D1 said watching on a TV is the ideal situation,
Interviewer: You seem to watch on your phone quite a even though he did not own one himself:
lot. Is that your preferred device? Interviewer: In an ideal world what would you choose to
F1: Yeah, I mean that’s when I’m at home. I think when watch on?
I’m [at work] I use my laptop.
D1: A really cool and expensive and nice TV.
Interviewer: So what do you like about the phone for
watching stuff on? Interviewer: And why is that?
F1: That it’s just more immediate. D1: Because the quality is quite nice, and if everything is
integrated with the streaming service and all that then...
However, mobile viewing was consistently seen as being un- lying down on the sofa is the best option.
favourable and was often avoided if possible. This was typi-
cally due to small screen sizes, as stated by household I: Viewing Locations
Interviewer: So do you ever watch on tablets or phones? We next consider the location where participants viewed con-
I2: No. tent. As shown in Figure 2, viewing occurred in 10 distinct
locations, with the living room and bedroom being the most
Interviewer: Never? Absolutely never? popular locations. We can again further collapse these loca-
I2: Never. tions into two distinct groups: watching in the home (living
Interviewer: Okay and why is that? room, bedroom, kitchen, etc.) and watching outside of the
I1: Screen is too small. home (workplace, public transport, public place). Households
reported more viewing sessions inside the home (M = 17.7, SD
Participant B1 spoke in disbelief that someone could watch = 8.9) than outside of the home (M = 2.1, SD = 2.2), and this
for long periods on a small screen: difference was significant, t(8)=4.82, p = .001. Moreover, four

72
 When considering how long participants viewed for, we found
that mean viewing session duration was 01:03:00 (SD =
00:55:56). A histogram of session durations can be seen in Fig-
ure 4. Of all the sessions, 122 were one hour or less (69%), and
158 sessions (89%) were two hours or less. Figure 5 shows a
detailed view of these sessions, where the most common dura-
tions is 30 minutes (often the length of one episode). Only 22
(12%) viewing sessions were over two hours. The longest ses-
sion was six hours, and the shortest two minutes. On average,
households reported longer viewing sessions on non-mobile
devices (M = 01:15:37, SD = 00:34:42) than on handheld mo-
bile devices (M = 00:38:24, SD = 00:18:58). However, this
difference was not significant, t(6) = 2.14, p = 0.076.

Figure 2: Distribution of viewing locations

Amount of Content Viewed
To better understand what was being watched during a session,
of the nine households never once reported watching content we also consider the amount of content that was watched. For
outside of the home. Most viewing sessions were reported to this analysis we consider each episode or separate video to a
have taken place in a single location; there were just five diary different item that is watched. Participants reported watching
entries (2.8%) in which participants reported moving between 481 items across 178 sessions; watching 2.7 items per session
two locations, and all of these were entirely inside the home. (SD = 2.7, range: 1–20). The largest number of items viewed
in a single session was 20 YouTube videos over 90 minutes.
Diary entries and interview data suggest that viewing location We found that households tended to watch more items on non-
was often not a concious choice, but a result of situational and mobile devices (M = 38.1, SD = 33.1) than on handheld mobile
contextual factors. Participant F1, living in a shared house devices (M = 15.3, SD = 23.4). However, this difference was
in London (where it is common to convert communal living again not significant, t(8) = 1.59, p = 0.15.
spaces into extra bedrooms), spoke about how she could not
watch in the living room:
On-demand Services Used
Interviewer: Why do you prefer to watching the bedroom Participants were also asked to record which on-demand ser-
than in the living room for instance? vices they used for viewing. They reported using 13 distinct
F1: Because I don’t have a living room. services. These are shown in Figure 6 along with the num-
ber of sessions they featured in. We divided these services
While small screens on mobile devices were often seen nega- into two categories: short-form, which consisted of YouTube,
tively, some participants spoke favourably about being able to Facebook, Lynda iOS app (a training course app), Vimeo,
view on public transport due to their portability, such as H1: WhatsApp, and The Guardian website (news); and long-form,
H1: You can use it on a plane. which consisted of Netflix, Raiplay (Italian on-demand ser-
vice), BBC iPlayer, unofficial streaming services, home record-
Interviewer: Why is that? ings, and Amazon Video. There was no difference in the num-
H1: Because you can just put it on the little table. ber of sessions reported by households for using either only
long-form services (M = 10.6 SD = 7.3) or only short-form ser-
Interviewer: Because it’s smaller? vices (M = 8.6, SD = 6.5), t(8) = 0.56, p = 0.59. Households
H1: It’s smaller. reported longer viewing sessions when sessions featured only
long-form services (M = 01:21:49, SD = 01:04:40) than with
Viewing Time of Day and Duration sessions containing only short-form services (M = 00:38:13,
We next consider the times at which participants watched SD = 00:30:40), and this difference was significant, t(8) = 3.75
during the day, and how long their viewing sessions lasted. p = 0.006.
A histogram of viewing start times can be seen in Figure 3.
It can be seen in the figure that late evening tended to be the Watching Alone and Watching Together
most popular time to start viewing, though lower levels of
We also explored whether people watched alone or with others
viewing also took place throughout the day, apart from in the
(i.e., co-viewing). Watching alone was more common than co-
very early hours of the morning. In terms of total viewing time,
viewing. In total, 135 sessions (75.8%) were watched alone,
105:08:00 (55.7%) of viewing took place in the evening period
and 43 (24.2%) by multiple people. In this context, co-viewing
between 18:00 and 00:00. It can also be seen in the figure that refers to more than one person actively watching.
viewing on handheld mobile devices was particularly popular
in the morning, and during late night and the early hours of the Motivations for watching alone were explored during the inter-
morning. There appears to be a noticeable transition from the views. A common theme was differing interests. For example,
pre-bed social ritual of watching on a TV to personal viewing household E (a cohabiting couple with 88.9% of their sessions
on mobile devices at bedtime. viewed alone), had very different tastes:

73
 Figure 3: Histogram of viewing start times

Figure 4: Histogram of viewing session durations

Figure 6: Popularity of different services

D1: We do really like very different things, and I think

the rare occasion we watch something together is... Well
actually, we do watch quite often The Big Bang Theory
during dinner. But it’s fifteen minutes and we watch the
same episodes all the time. So it’s more as kind of a
background thing... We tried watching other TV series
that we might enjoy watching together but those haven’t
existed to this point.
Interviewer: Okay, so you said she goes to sleep all the
time. Is it because she sleeps early and you go to bed
late? Is that a factor?
Figure 5: Histogram of viewing session durations for sessions D1: No no, it’s more that she only wants to watch what
with a duration of one hour or less she likes, and if I don’t adjust to it she really finds it really
boring and just falls asleep.
Interviewer: So what affects whether you watch together? Household A had a more even split of watching alone (58.33%)
Is it that you like different things, is it that you’re just versus together (41.67%), but still expressed different tastes
around at different times? which influenced whether or not they viewed together:
E1: Yeah I think I like watching it more than [E2] does, Interviewer: So, why don’t you want to watch what [A2]
and different things. I really enjoy watching House of wants to watch?
Cards whereas he’s more, I guess, logical with what he A1: Because it’s food programmes...
chooses to watch.
A2: Health...
Household D (53.3% of sessions viewed alone) also cited
A1: (Laughing) I can answer it myself. Yeah, it’s gener-
similar differences in personal tastes:
ally food programmes, health programmes...
Interviewer: And you said you don’t ever watch things Interviewer: Okay. And [A2], why don’t you want to
together? watch what [A1] wants to watch?
D1: Not really. [My girlfriend] falls asleep all the time. A2: I can’t watch another camper van conversion [on
Interviewer: [...] apart from that is there another reason? YouTube]!
Do you like different things? Interviewer: (Laughing) Okay.

74
 A2: And [A1]’s tutorials, he watches an awful lot of J1: I can always find something [more easily] on on-
tutorials, which wouldn’t interest me. demand, because on broadcast TV I am limited to [...]
forty different channels? And most of it is just reruns,
However, often watching alone was driven by situational fac- and on-demand I have the selection of fifteen-thousand
tors rather than conscious choice — sometimes people just videos or something.
happened to be alone when they watched. Participant F1
(90.91% of sessions watched alone), an international student, Participants also remarked on the quality of content available
discussed how she often watches alone when in the UK, but via on-demand services in comparison with broadcast TV,
with family when back in her home country: which was generally seen as similar if not better:
Interviewer: Do you normally watch alone then, when D1: I think I went into Netflix because of the catalogue
watching on-demand stuff? they had, and a couple of their own productions like
House of Cards, Orange Is the New Black, and the fact
F1: Yeah well, when I’m here, yes. ... If I go for holidays [that] they had a lot of stuff [...]. I tried it out and I
back home then I might do it with my, I don’t know, with enjoyed it.
my sister, or my mom.
However, participant J1 did note how broadcast TV was still
Participant G1 discussed how watching at work for a break useful for time-sensitive content:
meant they often watched alone:
J1: I think that quality is quite similar, I think. Broad-
G1: ...we like some similar shows and therefore we watch cast TV also has some more things to offer like as news
them together, but also because, I mean even for instance, coverage which you don’t get on demand.
[...] during lunch break if I’m alone, I watch something
and therefore I am alone! Negative Perceptions of On-Demand Viewing
Participant H1 also referenced her living situation, having re- While opinions were generally positive, participants noted
cently moved into a different household with new housemates: some negative aspects. Typically these were in relation to
watching large volumes of content, or at least the potential
H1: But maybe that’s just because of my living circum- to. Some participants spoke about being addicted to particular
stances. Before I used to just watch TV with other people. shows. Household A spoke about AMC’s Better Call Saul:
Interviewer: So which would you prefer? Or does it A1: Yeah, well we like to watch two or three at a time,
depend? don’t we?
H1: It depends, but I prefer to watch stuff with other A2: Yeah.
people I think. Interviewer: Why is that?
A1: We can’t stop watching them because they’re addic-
Positive Perceptions of On-Demand Video Services tive.
Participants generally had favourable opinions about on- A2: The trouble is, when you watch one that’s on for
demand services. One of the most obvious themes from the almost an hour... you just, you feel as though you’ve been
data was the benefits provided by these services in terms of short-changed, you need to watch some more.
freedom, convenience, and choice for viewers. The results
presented above show this clearly — participants watched in a A1: Yeah, you’re drawn in aren’t you? You just want to
variety of locations, at different times and on different devices. watch more.
They also spoke about this in interviews, such as the following Participant F1 also spoke about a compulsion to watch:
quote from participant A1:
F1: It becomes a bit addictive now. At least on the TV
A1: You can choose when you watch it then can’t you? if you missed an episode you will be like "Oh okay, I’ll
You know you don’t have to say "ooh it’s on at 8 o’clock just watch it next week" and then you will do other stuff.
tonight, we’ve got to be there for 8 o’clock". If we watch But now, I have this theory about habits. Because I can
it on-demand you can think "I’ll watch it at 10 o’clock if watch whenever I want, and it’s the sort of thing where I
I want". need to be watching now to go to sleep.
A number of households spoke about the catalogue of content Other participants also spoke about becoming hooked on a
available, which can make it easy to find something to watch: show, and how certain services made it very easy to watch
E1: I think I watched the first thirty minutes of it but another episode:
didn’t really... And that’s the thing about on-demand — if E2: Normally when you watch something you have to say
you don’t really like it, you can just find something else. "Should we see one more?". Then we would actually take
F1: It’s been so long since I watched the TV that I don’t an active choice to press next button, but Netflix there’s
even remember how it is that you have to wait every week like five seconds countdown. So often we [think] "Should
for a new episode or whatever, for the series, right? So we see one more?", "Hmm, I don’t really know" and then,
now I just watch them whenever I want, whenever I have the intro screen is on and Netflix started.
the time. E1: It made the choice for us.

75
 E2: Yeah, I think if it didn’t start automatically and we when pushed to define binge watching, few participants had
actually had to push the button, then I think we would a clear idea of what binge watching was. Some would define
talk. I probably would talk about if we should see one it as being based on the number of episodes of a show that
more, because now it was the fourth in a row. was watched, e.g. participant G1 defined it as three or more
episodes, but only when watching TV shows:
Participants also spoke about trying to control their viewing to
ensure they didn’t spend more time watching than intended: G1: I have always thought about it in terms of TV
shows [...]. So, watching many more than just one single
E1: You also want to see a lot of these like, Suits, or
episode, in one sitting.
whatever... We don’t want to get dragged into it because
I can’t get out of it (laughing). [...] I know myself too Interviewer: So how many episodes is it before you are
well that I’ll end up spending half a day there. binge-watching?
Participant H1 said something similar: G1: I would say from three.
Interviewer: So if I watch three five minute YouTube
H1: I didn’t want to start a series one time because I
videos, is that binge-watching?
knew I would just waste so much of my time watching it.
G1: Not exactly. My understanding was [...] that you are
This type of boundary setting was also mentioned by other watching episodes of 45 minutes each.
participants. Participant F1 thought that excessive viewing
might be having a detrimental effect on other areas of her life, Participant J1 also agreed with this:
and so spoke about creating hardware boundaries to combat it:
J1: I think binge watching should be sort of a TV series
F1: No I don’t have Netflix on my phone, and I don’t episode length. An hour, or 45 minutes, or 42 minutes...
want to put Netflix on my phone. and you watch more than two of those in a row.
Interviewer: Why is that? Others said it was based on the amount of overall time spent,
F1: Because at least with the tablet you know I leave it at such as participant D1:
home, and I know that I won’t use it unless I am at home Interviewer: How many episodes do you think is binge
at night. watching?
Interviewer: So that’s one way of setting a boundary?
R: Ooh, erm, anything that goes above four or five hours.
F1: Yes, I mean I always feel to set these boundaries and
they work, but the problem is for the last few months I’ve Interviewer: Okay, so it’s more about the time than the
become an addict to YouTube. I don’t think I was like number of episodes for you?
that last year. I was a bit more able to control myself. R: Yeah, because it’s not the same to watch a whole TV
Interviewer: How about watching outside of the office mini-series that has 10 episodes [that are] an hour and
and outside of home? So, maybe in a public place or 15 minutes each, [as it is to watch] ten episodes of The
while you were travelling for instance. Can you talk Big Bang Theory or Friends.
about if you did any of that? Participant I2 also agreed with this, specifically noting how
F1: No, because so first of all I don’t have enough data the number of episodes was inconsequential. He also seems to
to watch videos, and I also deliberately don’t pay more think that it is possible binge watch shorter content:
[...] so that I can restrict myself from watching videos,
I2: Well, the [time and number of episodes] are synony-
because otherwise I would just be watching everything.
mous, right? So if the programme was 10 minutes per
Participants also said how watching too much content often episode, then I would go through [many] more episodes
meant they wasted time or ended up going to bed too late: probably to achieve the same amount of time.
A2: I think sometimes we normally stay up a bit late with Household B also thought it was based on the amount of time
on-demand. spent, but disagreed about the actual definition.
H1: Um, well, because then I’ll watch maybe three Interviewer: So how would you define it? Is it the number
episodes in an evening...well, on a bad evening or like I’ll of episodes or is it the amount of time that you watch?
watch two and the next evening I’ll watch two. If there B2: The amount of time. [...]
are lots of episodes in the series then that’s a big waste
of time. Interviewer: So, how many episodes would have to watch
and how long would you have to watch for, for it to be
Binge Watching binge watching?
Discussions of consuming too much content often brought B3: I don’t know. I guess if you sit there, waste your
the subject specifically to binge watching, which was dis- whole night. [...]
cussed with all of the participants. Most of the participants
were familiar with this behaviour and said they participated B3: Yeah, I’ve never thought about it before, so I don’t
in it themselves. It seemed that this phenomenon could be know. Um, four or five hours I guess. [...]
thought of as a particular type of excessive viewing. However, B2: I’d go for three.

76
Such disagreement as to what constitutes a televisual binge viewing screen. However, changes as a result of new technol-
was also present in other households, such as in household A: ogy can also be seen. YouTube was the most common viewing
platform, showing how shortform content has become popular.
A1: Didn’t we watch three [episodes] in the last couple We also found that a third of viewing happened on a mobile
of weeks? We watched three [episodes]. device, and instances of very long viewing sessions.
A2: Oh three... yeah, but I wouldn’t say that constituted
binge watching, but maybe it does. When considering viewing screens, we can see that the laptop
was slightly more popular as a viewing device than the tele-
R1: I think three is, yeah. Three is, I would say, yeah. vision, which may not be possible without the cross-device
R2: I don’t know... but yeah we did watch three. availability of on-demand services. This could be due to the
Interviewer: So would you say that that’s binge watch- ease of access to different services via the internet, as well as
ing? the balance of screen size and portability that laptops provide.
However, for the purposes of viewing they function similarly
R2: I wouldn’t say that’s binge watching. to a TV — a fairly large screen that can be placed in a comfort-
Interviewer: Why not? able location, with the ability to watch with others. Though
R2: I don’t think there are enough episodes there. most viewing occurred on larger screens, a third of viewing
sessions were on handheld mobile devices. This was gener-
Interviewer: Okay. ally seen as unfavourable, and mostly seemed to be down to
R1: I would say... I think more than two is binge watch- necessity — in interviews, participants expressed their dislike
ing. for viewing on mobile devices, citing the small screen as a
reason. This agrees with previous work showing how viewing
Unlike some, H1 did not think that episodes necessarily had on small screens can lead to a reduced viewing experience
to be watched back-to-back or even on the same day: [15]. However, participants said that they would watch on a
H1: I think it’s watching multiple episodes compulsively. mobile device if no other device were available (e.g. when
[...] it could be one episode but you watch an episode per travelling). Most said they preferred to watch on a TV, due
evening or it could be within a shorter space of time... to large screen size and comfortable seating typically found
nearby. Individual differences were evident however, with
Participant J1 also suggested that watching one episode per some participants entirely discounting watching content on
evening could be binge watching, but was not entirely sure: phones, and others sometimes preferring it.
J1: Maybe it is... maybe seven episodes in seven days is We found that viewing device often depended on people’s
binge watching [...] it’s difficult to say. I think... like in a locations. The majority of viewing (89.9%) took place inside
short period of time, watching something that was made the home, and the living room was the most popular location.
for once a week maybe. It is perhaps then not surprising that people tended to watch
Participant I1 thought binge watching was more related to on larger display TVs and laptops rather than smaller mobile
viewing intentions: screens when in the living room. Mobile devices tended to be
used in the home in locations where there may not be access
I2: It depends, yes, because if I’m supposed to work to a TV, such as the kitchen or bedroom. Our participants did
and I tell myself, "Okay, one. One video" and then I report watching on mobile devices when outside of the home,
end up watching six, then it’s kind of binge watching too, particularly when travelling and commuting to work. In recent
because I was supposed to take just a 10 minute break. years the lower cost of mobile data has made easier to watch
I: So does it depend on what you are supposed to be on-demand services on the move. These instances of mobile
doing, for you? viewing tended to be during longer journeys, possibly because
R2: Yeah, I guess what the intention was. If I really it allows for an entire episode of content to be watched.
just want a fifteen-minute break and I end up, you know, Participants spoke very favourably about on demand services,
watching something for thirty minutes, then I kind of especially about how they have allowed them more freedom
escalated there, so in a way that would be binge watching. and choice than broadcast TV. However, a study by Vanat-
If it’s a lazy Saturday afternoon and it’s raining and I tenhoven and Geerts [20] found that some consumers found
end up watching three or four episodes, then yeah I think the amount of choice available to be an annoyance, especially
four or five is turning into binge watching, but otherwise with regard to the number of different services available. In-
if I have the time and nothing else to do... terestingly, we did not find this sentiment in any of our data.

DISCUSSION This freedom also allowed participants to select content that

The findings of this study show that although on-demand video matched their personal tastes. Our interview data revealed
platforms have the potential to change viewing behaviour, that these differing tastes among household members could
viewers still often conform to traditional viewing habits. For lead to people choosing to view alone, which was reflected in
instance, in terms of viewing time, most viewing occurred the diary data showing that 75.8% of sessions were watched
during the evening "prime time" slot. Furthermore, the most alone. This shows a stark turnaround of events when compared
common session duration was 30 minutes, typically the length with an observational study by Saxbe et al. [17], who found
of one episode of content. The TV was also still a popular that watching TV with at least one other person happened

77
for 61% of the time, and that the TV provided a platform for While we took effort to recruit participants of various ages and
togetherness in the household. While participant D1 said he living in different parts of the UK, most of our participants
and his partner generally watched different content in different were London-based millennials without children. This bias in
rooms due to differing tastes, participant A1 described how she the sample may have affected our results. For instance, some
and her partner used tablets and earphones to watch different participants lived in shared housing without a communal living
content, but still be in the same room together. This agrees room or TV. In place of this, viewing occurred on laptops and
with Ofcom’s findings [12] who found that people often turn tablets in bedrooms. Considering millennials’ typically high
to on-demand services for some "alone time". level of interaction with technology, we might have expected
more activity that differs from traditional notions of TV view-
Although we observed that only a quarter of sessions were ing. This may have seen an increase if our sample featured
co-viewed, we know from previous research that viewers value
more teenagers and children. Viewing mainly in the evening
the way new broadcasting technologies can enhance the social
is perhaps to be expected, as our sample was mostly adults in
aspect of viewing [1, 7]. While watching alone was more
full-time employment. However, there was a steady amount of
common than co-viewing, it may be that the sessions watched
daytime viewing, resulting from one household with children
alone were driven by other latent social factors, such as being being at home and people viewing during work breaks.
able to discuss the show with friends. Finally, it could be
that co-viewing and other social factors work differently in Our sample consisted of 20 individuals from nine households.
different household configurations, e.g. we observed that This could be argued to be a small sample size, however it is
the households with the highest percentage of co-viewing similar to that of comparable studies (e.g. [13, 1, 20, 9]). It
were household A, an older couple (42.1% of sessions co- also reflects the challenges of conducting this type of research,
viewed) and Household B, a family (41.7% of sessions co- where prolonged studies with involved tasks for participants
viewed), while the household with the lowest percentage was can deter participation, even when well compensated.This may
household H, two young professional cohabiting friends (6.7% have influenced the results of the statistical tests we employed
of sessions co-viewed). We cannot speculate beyond this with — while the differences between means were often quite large,
our data, but it would be an interesting focus of future research. the high variance and low sample size did return some non-
significant results. Notwithstanding, as the present study is
However, participants were often wary of the way instant qualitative in nature, we argue that the sample size is sufficient
access to large amounts of content could mean watching for to illuminate many of the behaviours surrounding on-demand
long periods. This led to some participants creating boundaries
and mobile viewing, especially given the study duration.
to prevent this behaviour, either by simply not starting to watch
a new show, or by restricting viewing in some other way, e.g. A limitation of the diary study method is that some participants
not installing Netflix on their phone. While it may in the may not have recorded everything they watched. During inter-
interests of service providers to make it as easy as possible to views some participants did remark that they sometimes did
view large volumes of content for revenue and engagement not record very short viewing sessions (e.g., a short Facebook
purposes, this was often troubling to our participants, some of video) because of the effort involved. However, this was fairly
whom commented that Netflix "made the choice for [them]" uncommon, with most participants saying they recorded the
when deciding whether to watch another episode. As such, the vast majority of content they watched.
introduction of small "design frictions" to combat automatic
behaviours could lead to a better user experience [4], either by CONCLUSION
design or manually by the users themselves. This paper extends our understanding of how on-demand view-
Discussing consuming large amounts of content typically led ing occurs in daily life. The results of a diary study show that
to talking about binge watching, which most of the partic- this technology leads to new behaviours such as mobile view-
ipants said they participated in. However, when pushed to ing, viewing for long periods, and consuming shortform con-
define what binge watching was, there were widely different tent. However, our sample still often conformed to traditional
responses and definitions often seemed to change depending viewing habits. Viewing was mostly in the evening on a large
on the context. This is reflected in other studies, where binge screen, though this sometimes happened in new ways, such
watching is defined differently by different authors. For ex- as by using a laptop. While mobile viewing did account for a
ample, some participants said it was watching two or more third of all viewing sessions, in general this was seen as less
episodes in a row (as in [14, 12]), and said three or more favourable than watching on a large screen. Typically, mobile
episodes in a row (as in [5, 21]). Others said it was not so viewing seemed to occur for contextual reasons, such as being
much the number of episodes watched but the total time spent a practical device to use while travelling, or wanting to watch
watching, while others said it was a combination of these two content privately when in the presence of others. We also
features. Others said that it depended on their intentions when found that viewing alone was far more common than viewing
they started to watch. In summary, different people seemed with other people. Participants had largely positive opinions
to have different ideas of what binge watching is, and this about on-demand video services, but generally seemed to be
disagreement reflects the diversity of definitions that appear in wary about the ability watch for long periods and the impact it
the literature on this topic. Such varying definitions suggest could have on other areas of their lives.
that it could defined on a scale, and vary with context and type
of content, as suggested by Trouleau et al. [19]. ACKNOWLEDGEMENTS
This research was supported by EPSRC grant EP/G037159/1.

78
REFERENCES Statistical Characterization and Consumption Patterns
1. Louise Barkhuus and Barry Brown. 2009. Unpacking the Identification on a Production Service. Multimedia
Television: User Practices around a Changing Systems 23, 5 (01 Oct 2017), 563–581. DOI:
Technology. ACM Transactions on Computer-Human http://dx.doi.org/10.1007/s00530-016-0516-7
Interaction (TOCHI) 16, 3, Article 15 (Sept. 2009), 22 11. Ofcom. 2016. Linear Vs. Non-linear Viewing: A
pages. DOI:http://dx.doi.org/10.1145/1592440.1592444 Qualitative Investigation Exploring Viewers’ Behaviour
2. Rhiannon Bury and Johnson Li. 2015. Is It Live or Is It and Attitudes Towards Using Different TV Platforms and
Timeshifted, Streamed or Downloaded? Watching Services Providers. (2016). http://www.ofcom.org.uk/
Television in the Era of Multiple Screens. New Media & research-and-data/tv-radio-and-on-demand/tv-research/
Society 17, 4 (2015), 592–610. DOI: linear-vs-non-linear-viewing
http://dx.doi.org/10.1177/1461444813508368 12. Ofcom. 2017. The Communications Market Report 2017.
3. The Nielsen Company. 2016. On-demand Demographics: (2017). http://www.ofcom.org.uk/__data/assets/pdf_
VOD Viewing Across Generations. (2016). file/0017/105074/cmr-2017-uk.pdf
http://www.nielsen.com/uk/en/insights/news/2016/
13. Kenton O’Hara, April Slayden Mitchell, and Alex
on-demand-demographics-vod-viewing-across-generations.
Vorbau. 2007. Consuming Video on Mobile Devices. In
html
Proceedings of the SIGCHI Conference on Human
4. Anna L. Cox, Sandy J.J. Gould, Marta E. Cecchinato, Factors in Computing Systems (CHI ’07). ACM, New
Ioanna Iacovides, and Ian Renfree. 2016. Design York, NY, USA, 857–866. DOI:
Frictions for Mindful Interactions: The Case for http://dx.doi.org/10.1145/1240624.1240754
Microboundaries. In Proceedings of the 2016 CHI 14. Matthew Pittman and Kim Sheehan. 2015. Sprinting a
Conference Extended Abstracts on Human Factors in Media Marathon: Uses and Gratifications of
Computing Systems (CHI EA ’16). ACM, New York, NY, Binge-watching Television through Netflix. First Monday
USA, 1389–1397. DOI: 20, 10 (2015). DOI:
http://dx.doi.org/10.1145/2851581.2892410
http://dx.doi.org/10.5210/fm.v20i10.6138
5. Dimph de Feijter, Vassilis-Javed Khan, and Marnix van 15. Jacob M. Rigby, Duncan P. Brumby, Anna L. Cox, and
Gisbergen. 2016. Confessions of a ‘Guilty’ Couch Potato Sandy J.J. Gould. 2016. Watching Movies on Netflix:
Understanding and Using Context to Optimize Investigating the Effect of Screen Size on Viewer
Binge-watching Behavior. In Proceedings of the ACM Immersion. In Proceedings of the 18th International
International Conference on Interactive Experiences for Conference on Human-Computer Interaction with Mobile
TV and Online Video (TVX ’16). ACM, New York, NY, Devices and Services Adjunct (MobileHCI ’16). ACM,
USA, 59–67. DOI: New York, NY, USA, 714–721. DOI:
http://dx.doi.org/10.1145/2932206.2932216
http://dx.doi.org/10.1145/2957265.2961843
6. David Gauntlett and Annette Hill. 2002. TV Living:
16. Jacob M. Rigby, Duncan P. Brumby, Anna L. Cox, and
Television, Culture and Everyday Life. Routledge.
Sandy J.J. Gould. 2018. Old Habits Die Hard: A Diary
7. Lilly Irani, Robin Jeffries, and Andrea Knight. 2010. Study of On-demand Video Viewing. In Proceedings of
Rhythms and plasticity: television temporality at home. the 2018 CHI Conference Extended Abstracts on Human
Personal and Ubiquitous Computing 14, 7 (01 Oct 2010), Factors in Computing Systems (CHI EA ’18). ACM, New
621–632. DOI: York, NY, USA. DOI:
http://dx.doi.org/10.1007/s00779-009-0280-1 http://dx.doi.org/10.1145/3170427.3188665

8. Robert J. Logan, Sheila Augaitis, Robert H. Miller, and 17. Darby Saxbe, Anthony Graesch, and Marie Alvik. 2011.
Keith Wehmeyer. 1995. Living Room Culture - an Television As a Social or Solo Activity: Understanding
Anthropological Study of Television Usage Behaviors. In Families’ Everyday Television Viewing Patterns.
Proceedings of the Human Factors and Ergonomics Communication Research Reports 28, 2 (2011), 180–189.
Society Annual Meeting, Vol. 39. SAGE Publications
18. Alex Taylor and Richard Harper. 2003. Switching off to
Sage CA: Los Angeles, CA, 326–330. DOI:
Switch On. In Inside the Smart Home, Richard Harper
http://dx.doi.org/10.1177/154193129503900507
(Ed.). Springer London, London, Chapter 7, 115–126.
9. Jennifer McNally and Beth Harrington. 2017. How DOI:http://dx.doi.org/10.1007/1-85233-854-7_7
Millennials and Teens Consume Mobile Video. In
Proceedings of the 2017 ACM International Conference 19. William Trouleau, Azin Ashkan, Weicong Ding, and
on Interactive Experiences for TV and Online Video (TVX Brian Eriksson. 2016. Just One More: Modeling Binge
’17). ACM, New York, NY, USA, 31–39. DOI: Watching Behavior. In Proceedings of the 22nd ACM
http://dx.doi.org/10.1145/3077548.3077555
SIGKDD International Conference on Knowledge
Discovery and Data Mining (KDD ’16). ACM, New York,
10. João Nogueira, Lucas Guardalben, Bernardo Cardoso, NY, USA, 1215–1224. DOI:
and Susana Sargento. 2017. Catch-up TV Analytics: http://dx.doi.org/10.1145/2939672.2939792

79
20. Jeroen Vanattenhoven and David Geerts. 2015. Broadcast, 21. Emily Walton-Pattison, Stephan U. Dombrowski, and
Video-on-demand, and Other Ways to Watch Television Justin Presseau. 2016. ‘Just One More Episode’:
Content: A Household Perspective. In Proceedings of the Frequency and Theoretical Correlates of Television Binge
ACM International Conference on Interactive Watching. Journal of Health Psychology 23, 1 (2016),
Experiences for TV and Online Video (TVX ’15). ACM, 17–24. DOI:http://dx.doi.org/10.1177/1359105316643379
New York, NY, USA, 73–82. DOI:
http://dx.doi.org/10.1145/2745197.2745208

80