Contextual patterns in mobile service...

ORIGINAL ARTICLE

Contextual patterns in mobile service usage

Hannu Verkasalo

Received: 6 March 2007 / Accepted: 5 February 2008 / Published online: 4 March 2008

� Springer-Verlag London Limited 2008

Abstract Mobile services differ from other services

because of their temporal and spatial attributes. Mobile

services additionally differ from each other in their value-

added to the end-user. Some services—such as emailing

and voice—are more business oriented. On the other hand,

various free-time oriented services are provided in new

smartphones, such as imaging and music playback. The

present paper studies how mobile services are used in

different contexts. For this, the paper develops a special-

ized algorithm that can be used with handset-based usage

data acquired straight from end-users in an established

panel study process. Educated guesses can be drawn on the

user context based on the developed algorithm. In the

present exercise usage contexts were divided into home,

office and ‘‘on the move’’. The algorithm is used with

exemplary data from Finland and the UK covering 324

consumers in 2006. More than 70% of contextual use cases

are correctly classified based on raw data. According to

exemplary results particularly multimedia services are used

‘‘on the move’’, whereas legacy mobile services experience

more evenly distributed usage across all contexts. The

algorithm that identifies context based on raw data provides

a new angle to mobile end-user research. In the future, the

accuracy of the algorithm will be improved with the inte-

gration of seamless cell-id logging and GPS data.

Keywords Context-oriented services �Ubiquitous computing � Mobile services

1 Introduction

The mobile services industry is growing fast. According to

Nokia’s estimations, the number of cellular subscriptions is

likely to surpass three billion in 2008 (Nokia 2005). Many

people get actually their first Internet use experience with

mobile handsets. The portfolio of mobile services expands

also in other dimensions, for example, new mobile multi-

media and navigation services emerge on the market. The

most advanced pocket-sized mobile handsets (i.e. smart-

phones, converged devices) support many functions from

multimedia to personal information management. Person-

to-person services, however, still remain as the most

important service class. According to recent results from a

Finnish smartphone study in 2006 about 60% of smart-

phone usage time is still allocated on person-to-person

oriented services. This is down from 70% in 2005 (Ver-

kasalo 2007b).

As the evolution of the mobile services industry pro-

ceeds further, more value can be extracted from

understanding the usage patterns of individual customers.

New smartphone devices provide many different kinds of

value-added to the end-user. The understanding of new

services holds value because these new services might have

disruptive effects (Christensen 1997; Barsi 2002) on the

whole future of the mobile services industry (Verkasalo

2007a). The concept of ubiquitous computing in the mobile

domain is used while referring to today’s heterogeneous

mobile services. This concept refers to a trend in which

mobile services integrate with the environment, improving

the contextual value through, e.g., entertainment, infor-

mation or communication (see e.g. Ojala et al. 2003;

Abowd et al. 1997). To give an example, music playback

services provide lots of value-added when people are

commuting, but there is less value-added from such

H. Verkasalo (&)

Helsinki University of Technology, Helsinki, Finland

e-mail: [email protected]

123

Pers Ubiquit Comput (2009) 13:331–342

DOI 10.1007/s00779-008-0197-0

services at home because substitute devices (home stereo

sets) exist. Similarly the alarm clock service might be

practical at home, but there are no valuable use cases for

such service at office. In the future smartphones can assist

in various kinds of situations such as doing shopping

(payments, retrieval of product information), navigating to

a particular point of interest, retrieving weather forecasts,

confirming the airline check-in, etc. Context is much more

than the mere location; it is something that relates to all

situational factors around the user in a particular use case

(Kaasinen 2003).

This paper discusses the contextual dimension of mobile

services. Then a specialized context algorithm is developed

for a handset-based smartphone service research platform

(Verkasalo and Hammainen 2007). The algorithm analyses

usage-level data and divides individual customers’ usage

into three separate contexts (home, office, ‘‘on the move’’).

In the final part of the paper exemplary research results of

the algorithm are presented based on panel studies arranged

in the UK and Finland in fall 2006. This prototype algo-

rithm holds lots of future potential, particularly when

integrating it with seamless cell-id logging and utilization

of GPS data from handsets. No earlier approaches in

modeling the context of actual end-users exist in this scale.

Based on the outputs of the algorithm usage-level behavior

of smartphone users can be analyzed in great detail. In the

future, the algorithm will contribute to various disciplines

from end-user marketing to new service road tests and from

behavioral sciences to new business development.

2 Mobile services and context

Legacy mobile services include voice and text messaging.

These services have been embedded basically in all mobile

phones as we know them today (pocket-sized handheld

devices with digital cellular connectivity). Smartphones,

which combine person-to-person services with computer-

like applications such as document viewers, Internet

browsing and multimedia, have stretched the functionality

of these handheld devices. At the same time they have

made mobile phones even more all-around devices. Not

only do we see both hedonic and utilitarian mobile services

today, but also functionally these services provide many

kinds of value-added to the end-user.

A service is generally considered as a non-material

equivalent of a good. According to Kotler and Armstrong

(1996) services are ‘‘activities or benefits offered for sale

that are essentially intangible and do not result in the

ownership of anything’’. Electronic services are thereby

services, which involve electronic transactions. A mobile

service, further, is defined in this paper essentially as a

service that is consumed by the end-user with a mobile

handset. In this definition, a mobile handset means a

pocket-sized, handheld terminal which has at least a cel-

lular radio interface implemented.

In this paper, the focus is particularly on core mobile

services that provide value-added straight to the end-user.

Core services are built on top of enabler or supporting

services. Enabler and supporting services are needed in

implementing the core service. For example, a generic

packet data interface is needed in implementing video

streaming applications that utilize Internet protocols. In

this example, the packet data interface is a service enabler

whereas the video streaming application itself is the core

service. Different kinds of mobile services have different

characteristics, and therefore they should be managed and

deployed in different ways (Balasubramanian et al. 2002;

Heinonen and Andersson 2003). From a contextual point of

view, for example, multimedia services should be easy to

launch when walking on the street, mobile browsers should

be optimized for small displays and physical restrictions,

and future voice services should leverage, e.g. WiFi net-

works (with mobile VoIP) at home/office and cellular

network access (with legacy circuit-switched voice) when

outside of home/office (mainly because of more efficient

use of network resources).

Applications are here seen either as network applica-

tions implementing the service or then as stand-alone

applications in the handset (i.e. terminal applications).

Application is, therefore, a more technical term referring to

the solution itself, whereas service is better associated with

the value-added to the end-user. In this paper, all solutions

providing value-added to the end-user through mobile

handsets are collectively called as mobile services.

Mobile services have some special characteristics in

comparison to other types of services. The key differences

are related to the spatial and temporal components of ser-

vice usage. If you want to meet a bank teller, you have to

visit the bank (location) at a certain appointment (time).

These restrictions are sometimes present even when using

some electronic services, such as the Internet with fixed

access (i.e. wired access). Even though the fixed Internet

provides e.g. online banking services 24 h a day (thus

overcoming the problem of temporal availability), you still

need a computer with fixed access (which causes problems

in the spatial domain). Mobile services, used with handheld

mobile devices, overcome both spatial and temporal con-

straints (Heinonen and Pura 2006). One more interesting

dimension of mobile services is the potential for individual

personalization of service offerings (Clarke and Flaherty

2003). According to Kaasinen (2003) mobile services

should adapt in various dimensions (e.g. location and

context) to user preferences.

Many approaches to classify mobile services exist. To

mention some of the alternative approaches, the classification

332 Pers Ubiquit Comput (2009) 13:331–342

123

can originate from the end-user (subscribers, service users),

terminal (mobile phone, PDA, laptop computer, desktop

computer), network (operator, network technologies, serv-

ers) and academic research (diffusion/adoption research,

market studies, technical studies) points of view.

Verkasalo, Kivi and Smura (2007) identify three streams

of literature with regards to the mobile service classifica-

tion. First, the technical/business perspective exists as quite

a logical way to formulate a classification (see e.g. Ange-

hrn 1997; Basaure 2004; Cerqueira et al. 2005; ECOSYS

2004; Meuter et al. 2000; Mitchell and Whitmore 2003;

UMTS Forum 2001; Vesa 2005) framework. Technical and

business approaches are often tied together, and the per-

spective is external to the service user.

Secondly, the end-user perspective in classifying services

remains as the most common approach. This approach is

applied at least in Anckar and D’Incau (2002), Chande

(2005), Dabholkar (1996), Giaglis et al. (2003), Hyvonen

and Repo (2005), FMTC (2004), Jarvenpaa and Lang

(2005), Kaasinen (2005), Nysveen et al. (2005), O’Loughlin

(2000), Sullivan et al. (2005), Vanjoki (2003), Velez and

Correia (2002), Verkasalo and Hammainen (2006), Wil-

liamson (2003), Korhonen (2003), and Yoon et al. (2003).

The end-user perspective is much internal, looking at service

offerings from the end-user point of view.

Thirdly, an emerging stream of literature utilizes holis-

tically many softer research disciplines. This stream deals

with, e.g., context-specific use cases and special ways of

implementing particularly mobile services. Some of the

papers utilizing this approach include ARC Group (1999);

Balasubramanian et al. (2002); Heinonen and Pura (2006);

Heinonen (2004); Pura and Brush (2005); Van der Hejden

et al. (2005). This third approach increasingly combines

softer academic disciplines such as sociology or psychol-

ogy with older user-centric approaches of mobile service

literature. Heinonen and Pura (2006) claimed that the

spatial and temporal (space and time) dimensions can be

used in analyzing the characteristics of a mobile service

(see also Balasubramanian et al. 2002). The spatial and

temporal contexts can also be extended to include the

social setting (see e.g. Okazaki 2005). Furthermore, it is

suggested that the relationship between a customer and

service provider (e.g. the extent of personalization) should

be used in the service classification (Heinonen and Pura

2006). All these relate more or less to the sociological,

psychological or contextual service analysis. The context

perspective, that is the focus of this paper, is not far from

the end-user perspective in reality. This is because the

context is still perceived by the end-user (Kaasinen 2003),

and end-users should never be overlooked in mobile ser-

vice analyses (Holtel 2006).

The micro-level mobile service perspective (explained

above), focusing particularly on the contextual dimension

of service analysis, is now taken as the principal perspec-

tive of this paper. As mobile devices are carried basically

everywhere by end-users, they are therefore devices that

perfectly reflect the user’s context. Context is much more

than the location, including also the social setting and sit-

uational factors. Kaasinen (2003) argues that besides

location the other elements of context, however, are diffi-

cult to identify and measure. As the mobile services

industry and enhanced software platforms provide quite a

heterogenic portfolio of services today, it makes sense to

study these services not only from the user demographics

(Verkasalo 2005), but also from the contextual usage point

of view. The present research approach provides a signif-

icant contribution in understanding end-users by tackling

the contextual dimension with specific data mining tools.

Earlier studies that have discussed the modeling of con-

text and location with cell-ids include, e.g., Carmichael et al.

(2005), Trevisani and Vitaletti (2004) and Li et al. (2006).

The context algorithm introduced in this paper is used with a

recently developed handset-based mobile service research

platform (Verkasalo and Hammainen 2007). A handset-

based client extracts data on location (cell-ids) and time of

use. Based on this, educated guesses can be made on the

context of usage. In this paper, exemplary aggregated study

results are presented to demonstrate how the algorithm

could be used in large end-user panels with the associated

research platform. This research paper is pioneering in the

sense that the handset-based research platform has not been

earlier used in context research. Potential venues benefiting

from this research approach include

• Development of better UIs for different contextual

situations

• Adaptation of smartphone functionalities to different

contexts

• Potential improvements in presence services and

performance

• Extensions of mobile subscriber segmentation studies

• Analysis of social networks and micro-level behavior

of subscribers

• Preparation of techno-economic comparisons between,

e.g., competing radio access technologies (Smura 2006)

3 Identifying context from handset-based data

The solution to the identification of the context is explained

in this section. The presented algorithm is developed for a

specialized mobile service research platform. Handset-

based smartphone usage data is first acquired through a

novel service research platform. The platform consists of a

smartphone client running on subscribers’ Nokia S60

smartphones that acquires device actions and writes them

Pers Ubiquit Comput (2009) 13:331–342 333

123

to memory. Then this data is transmitted to centralized

servers for further analysis. For an overview of the process,

refer to Verkasalo and Hammainen (2007). The raw data

including all user actions with the smartphone over a panel

project of 2–3 months by several hundred subscribers is

used together with accurate hourly cell-if logs in deter-

mining the context of usage and then further in analyzing

usage-level differences between different contexts.

The approach is not unique, as Raento et al. (2005)

developed tools for identifying the context, too. Their cli-

ent automatically identified clusters of locations and

provided some artificial intelligence in making guesses on

users’ context and presence information. This client has

been used by Eagle (2005), too. The principal difference in

the approach of this paper is the data acquisition system. In

this paper, the data is centrally managed, and controlled

panel studies are analyzed in an aggregate fashion. In other

words, with the present approach much more data can be

collected and very accurate usage-level analyses between

contexts can be made. Whereas Raento et al.’s (2005) is a

more natural approach in building intelligence to the

handset, the approach of this paper better suits for large-

scale end-user research.

The input of the algorithm is a data file presenting each

user’s active cell-id (the unique cell-id identification code

to which the subscriber’s mobile phone is currently

hooked), logged once per hour for each panelist over the

whole panel period. Only this data file is required for the

proper functioning of the algorithm. Other data files rep-

resenting users’ smartphone actions (e.g. voice

communications, multimedia application usage, data ses-

sion details), equipped with time stamps, are later needed

in combining the contextual information with actual usage.

The stages of the context algorithm process are:

1. For each particular customer, the cumulative usage

(e.g. active hook-up hours) of each cell-id is calculated

over the whole panel period. Then the relative share of

usage for each of these cell-ids is calculated by

dividing the number of a particular cell’s active hours

by the total number of active hours spent in the panel.

2. For each cell-id achieving a relative share of usage

more than 10% (i.e. there is significant usage under a

particular cell) a hypothesis is made that the cell-id

must represent either home or office usage (both are

fixed locations with potential for a high number of

active usage hours).

3. The cells with at least a relative usage ratio of 10% are

analyzed in greater detail by weighting their hourly

and weekday-based distributions with corresponding

statistics (provided by Statistics Finland) indicating

people’s averaged location dynamics per weekday and

hour. Those cells that receive relatively more hits

during typical home context hours get more ‘‘home

points’’, and vice versa with office context.

4. ‘‘Home’’ and ‘‘office points’’ are summed up in order

to classify cells with relatively high cumulative usage

either into home or office contexts. If the cell received

more home points all usage taking place under that cell

is considered home usage, and vice versa for office.

5. All the cells receiving less than 10% of all cell-id hits

are considered as ‘‘on the move’’ context. The

assumption is thereby made that people do not spend

more than 10% of cumulative time in any of the ‘‘on

the move’’ cells.

6. All users for whom three separate contexts with

significant ([100 h) usage per context are identified

and included in the final dataset for usage-level

comparisons.

7. The outputs of the context algorithm are mapped

together with usage data in order to classify all other

user actions into one of the three contexts. Based on

this classification accurate usage-level comparisons

can then be made on the services under interest.

The context algorithm can be run on the whole aggre-

gated panel dataset. The algorithm is programmed into an

SPSS (2002) syntax file that is easy to apply for new panel

studies. The contribution of the algorithm is an automated

mapping of accurate logs on user location and timestamps

to statistics from Statistics Finland reflecting daily contex-

tual distributions of a high number of interviewed people.

The outputs of the algorithm are further easy to combine

with usage-level data. The major challenges include the

discontinuity of cell-id logging (only once per hour), net-

work conditions (change of the active cell-id because of

poor reception or network load even though the user does

not move) and overly mobile behavioral patterns (e.g.

people who do not have fixed daily office locations cannot

be easily analyzed). In the future, the context algorithm will

be improved by utilizing seamless cell-id logging (writing

logs every time the cell-id changes) and GPS chip data.

4 Empirical observations on contextual mobile service

usage

Based on the developed algorithm and available data,

demonstrative usage-level results of the context algorithm

are now presented. Though the panel studies (including

several hundred smartphone customers with the monitoring

software installed for ca. 2–3 months in a given market) are

geared towards early-adopter customers (see e.g. Rogers

1962), they nevertheless tell something about the most

advanced usage patterns that might evade to the mass-

market domain in the near future. This study sets out to


123

provide some exemplary results on contextual usage-level

behavior. Future studies with the improved version of the

algorithm should be used in providing robustness studies

with better accuracy.

A dataset from Finland and the UK is utilized as a case

example of the utilization of the context algorithm. All in

all, the original dataset included 861 users. All of these

users had generated at least 3 weeks of active smartphone

usage. After running the context algorithm over the whole

dataset, 324 smartphone users were identified for whom the

context algorithm accurately identified three separate

contexts. For many users the context algorithm did not find

a proper office context. This indicates that those users were

perhaps not working regularly (e.g. unemployed people or

students who do not visit even school context too often),

they were exceptionally mobile workers, or they work from

multiple locations. Secondary handset usage (which makes

the dataset discontinuous) or poor cell-id reception (i.e. lots

of cell-id jumps in office context) complicate the context

identification process, too. The remaining 324 users for

whom the context algorithm worked well are included in

the results below. These 324 users generated an average of

67 usage days during the panel study. In terms of demo-

graphics they were mostly male and 20–39 years old

people, resembling demographic distributions observed in

other studies (see e.g. Verkasalo 2007b).

Figure 8 in Appendix A shows exemplary outputs of the

context algorithm for one particular user. As can be seen,

the algorithm divides all observed usage into three distinct

classes based on the heuristics as explained in the technical

part of this paper. In panel-wide studies, however, all data

is aggregated together and typically it makes more sense to

report, e.g., the share of all panelists who have explored a

particular service in a certain context, or the cumulative

share of all outbound voice calls that took place in home

context. Figure 8 nevertheless indicates the level of accu-

racy (hourly contextual information for every single

panelist) the context algorithm can reach.

During the panel study a possibility emerged to push

pop-up questions to users’ handsets through the handset-

based research platform. An interactive question related to

the user’s context was asked from more than 100 customers

after a random packet data session taking place during the

panel; 87 answers were received indicating the users’ own

perception of their context (home, office, ‘‘on the move’’).

This data was linked to the outputs of the context algo-

rithm; 70% of cases were correctly classified, i.e. the user’s

opinion and the guess of the context algorithm were

aligned. This is quite a satisfying success rate, as the

context algorithm in its current form still has the limitation

of only an hourly accuracy (therefore, the active context

might have changed after the hourly cell-id logging, and

thus the pop-up question’s answer is time-wise separate

from the output of the context algorithm). Additional error

is caused by people having mixed opinions on the type of

context (i.e. the context algorithm names typical weekday

daytime contexts into ‘‘office’’, though the user might

actually visit, e.g., alternative offices and consider this as

‘‘on the move’’ usage).

Figure 9 in Appendix A presents the distribution of

smartphone usage activity. The most active users have about

14 active smartphone usage hours per day, meaning that they

have taken actions with the smartphone on average during

14 out of 24 h of the day. On the other hand, users with very

low usage activity exist (e.g., only a couple of unique hours

of active smartphone usage per day). Those for whom a lot of

‘‘on the move’’ usage was identified, generated more

smartphone usage hours on average. Thus, more mobile

subscribers use smartphones more actively. Smartphones

are still meant in the first place for mobile usage.

Based on the aggregate data, panelists on average were

located 49.9% of weekday time in home context, whereas

office and ‘‘on the move’’ contexts represented 23.5 and

26.6% of the time, respectively. Home context acquires the

most time because people typically not only hang around a

lot at home, but they also sleep at home (thus generating a

lot of inactive presence at home).

Figure 1 presents averaged weekday context distribu-

tion. As can be seen, office context acquires the most usage

between 8 a.m. and 3 p.m., reflecting typical office hours.

Some ‘‘on the move’’ usage can be identified in the

nighttime, reflecting, e.g., foreign trips, holidays, visits to

friends, etc.

Figure 2 presents context distribution over the days of

the week. Some differences can be identified between

weekdays. For example, people on average spend 24.3% of

daily time (*5.8 h) at office on Monday and only 21.3%

(*5.1 h) on Fridays. Taking into account that the dataset is

quite extensive, these differences suggest that people spend

more time at work on Mondays, whereas an average

workday on Fridays is shorter. Very little office usage can

be identified on weekends.

Figure 10 in Appendix A illustrates aggregated time

allocation results per context. Multimedia usage experi-

ences more usage ‘‘on the move’’ than in office or home

contexts, suggesting that people utilize smartphone camera

and music playback functionalities relatively more some-

where else than at home or office. Voice call usage receives

relatively more usage at home than at office or ‘‘on the

move’’. Taking into account that voice is perhaps the most

context-independent service (most people have already

substituted mobile phones for fixed-line telephones), this

observation is plausible.

Figure 3 presents that on average panelists use the

smartphone the most at home. A typical panelist spends 69

min per day by doing something with the handset. It was


123

earlier observed that 49.9% of all time is spent at home,

and Fig. 3 suggests that 51.8% of active smartphone usage

time (i.e. smartphone has been actively used) is spent in the

same context. The relative usage activity time (including

only active usage) follows the distribution of relative

context-specific time (including both inactive and active

usage). This is surprising, as one would assume that people

slept more at home than at office or ‘‘on the move’’. Per-

haps people still play around with smartphones quite a lot

at home, thus explaining the numbers of Fig. 3.

Figure 11 in Appendix A presents person-to-person

action distributions for weekdays. As can be seen, both

voice and SMS are all-around services. People utilize them

equally much in all contexts. SMS is a bit more home-

oriented service than voice, supporting earlier observations

suggesting that SMS is more free-time oriented. According

to earlier results SMS has caught relatively high usage in

the evening. Voice has been found more business-oriented

(Verkasalo 2005).

Figure 4 projects both mean and median durations for

inbound and outbound voice calls. In addition to normal

weekday contexts, also home and ‘‘on the move’’ weekend

contexts are included in this analysis. Voice calls tend to be

significantly longer at home, and this holds irrespective of

weekday or weekend. Perhaps in other contexts than home

there is less time available for casual chatting, whereas at

home people enjoy talking to friends for longer.

Figure 5 presents findings on more advanced services.

There are no significant differences between contexts when

comparing usage coverage numbers that reflect simply the

share of panelists who have at least once tried a particular

service in a given context. However, the fact that there is

less ‘‘explorative usage’’ in office context is clearly evident.

Figure 6 presents contextual usage patterns of particular

smartphone applications. Productivity applications such as

Weekday Hourly Location

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23

HourS

har

e o

f D

aily

Ho

urs

Sp

ent

Home Office On the move

Fig. 1 Weekday hourly

distribution of context

Weekly Location

10%

20%30%

40%

50%

60%

70%80%

90%

0%

100%

Tuesday Wednesday Thursday Friday SaturdayMonday Sunday

Day of Week

Sh

are

of

Wee

kly

Ho

urs

Sp

ent


Fig. 2 Daily distribution of

context

35.6

14.5

18.6

0

5

10

15

20

25

30

35

40


To

tal M

inu

tes

of

Usa

ge

/ Co

nte

xt /

Day

Fig. 3 Contextual smartphone usage minutes


123

Adobe Reader catch quite a lot of usage ‘‘on the move’’. In

office and home contexts people perhaps have an access to

laptops or desktop computers and there are therefore fewer

needs to use productivity applications with mobile handsets.

File manager is a kind of configuration application (used when,

e.g., moving files and installing applications), and this kind of

usage mostly takes place at home. Finally, Clock application is

almost solely used for setting up an alarm. No surprise, almost

all usage of Clock application takes place at home.

Figure 7 reflects contextual usage patterns of multi-

media services. As can be seen, almost all but gaming

usage takes place ‘‘on the move’’. Multimedia applications

are—similarly to productivity applications—of such kind

that the value-added to the end-user is realized ‘‘on the

move’’. At home or office substitutes for these services

(music playback) exist or then there is simply no need for

the service (not that many people capture photos at home).

Interestingly, there is a lot of gaming usage at home. This

supports earlier speculations indicating that people might

play around with mobile handsets quite a lot at home.

Figures 12 and 13 in Appendix A tell about mobile

Internet usage. In general all three most important Internet

service categories (browsing, email and instant messaging)

experience the most usage ‘‘on the move’’. This can be

once again reasoned by the fact that there are simply fewer

needs for mobile Internet usage at home or office, because

substitute devices that are easier to use (desktop/laptop

computers) are available in these contexts. Browsing ses-

sions are longest ‘‘on the move’’, reflecting that people

have more hedonic (‘‘killing time’’) needs for browsing in

this context. At home or office there is probably a relatively

small need and, therefore, browsing sessions tend to be

shorter there.

All in all, these brief examples present the kinds of

usage-level analyses that can be pursued with the context

algorithm and the supporting handset-based mobile service

research platform. There are various usage patterns evident

proving that mobile services provide different kinds of

contextual value-added. Future usage-level studies could

pursue the following topics:

Fig. 4 Voice call durations in

different contexts

49%

9%

49%

19%

32%

46%

5%

42%

13%

19%

8%

49%

22%

32%

55%

0%

10%

20%

30%

40%

50%

60%

Browsing WLAN Access OfflineMultimedia

Playback

Radio VideoRecording

Sh

are

of

Pan

elis

ts T

ried


Fig. 5 Advanced service usage

in different contexts


123

• Segmentation of users based on mobility patterns and

mobile service usage

• Visualization of mobile subscriber movement patterns

• Techno-economic calculations leveraging contextual

information (e.g., the potential of WiFi in substituting

for cellular access at home)

• Analyses of individual applications and their potential

for technical improvements (e.g., context adaptation)

The context algorithm together with the handset-based

end-user research platform provides unique data that is

difficult to obtain otherwise.

5 Conclusion

The present paper discussed first the special characteristics

of mobile services. Most importantly, mobile services can

be used from anywhere at any point of time. Mobile ser-

vices cover nowadays quite a portfolio of services from

hedonic to utilitarian ones. A need exists to empirically

model contextual mobile service usage.

A specialized context algorithm was developed to be

used together with a handset-based end-user research

platform. The new algorithm along with the associated

process of obtaining and analyzing huge amounts of usage-

level data automatically helps in assessing contextual

dimensions of mobile services. The new process was tested

with a case example including data from Finland and the

UK from fall 2006. According to the results, multimedia

and Internet services are used quite actively ‘‘on the

move’’, whereas legacy SMS and voice services experience

more evenly distributed usage among home, office and ‘‘on

the move’’ contexts. Different people have unique mobility

patterns, and therefore future end-user research should

78%

48%

29%

3%

22%

27%

19%

29%

44%

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%

Clock applicationusage

File managerapplication usage

Adobe Readerapplication usage

Share of All Actions

Home Office On The Move

Fig. 6 Productivity application

and clock usage in different

contexts

31%

29%

24%

44%

37%

22%

21%

22%

28%

17%

47%

49%

53%

28%

46%

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%

Offline multimedia player usage

Multimedia application usage:music/sounds

Radio application usage

Gaming usage

Camera usage



Fig. 7 Multimedia usage in

different contexts


123

focus on the micro-level of contextual service usage and

specialized segmentation studies. Other levers of the

developed algorithm are context-aware applications such

as presence solutions. The new algorithm and usage-level

data retrieval platform provide valuable information on the

behavior of mobile subscribers and service usage patterns.

This information might be useful among application

developers, strategic decision markets of mobile operators

and academic researchers.

Contextual analysis of end-user presence provides stra-

tegic insights. For example, handovers making seamless

transfer of voice calls from cellular to home WiFi networks

are nowadays possible with UMA technology. In order to

analyze network-level impacts of an introduction of WiFi/

cellular dual radio access, accurate data on consumer

contextual behavior is needed. Accurate analysis of mobile

services in different contexts also improves the under-

standing of service usage dynamics. This understanding is

needed in service commercialization and strategic posi-

tioning tasks.

The main venues for technical contribution are related to

an update of the algorithm with seamless cell-id logging

(continuous instead of hourly logging). In addition, future

handsets with embedded GPS chips could provide even

more accuracy to available location data. In particular, the

context could be identified with higher accuracy, and

moving buses can be separated from stable locations, e.g., a

cafeteria.

The heuristics presented in this paper did not identify the

context for everybody. The main source of error is too low a

frequency of cell-id logs and incapability to understand

geographical proximity (due to unavailability of GPS data

yet) and resulting inaccuracy in cumulating usage times per

context. With the planned improvements it is expected that

the presented heuristics can provide contribution to context-

oriented end-user research in the future. The exemplary

results reflecting differences in end-user behavior in dif-

ferent contexts will be further examined in the future. The

study platform will be improved to better account for both

explorative and confirmatory mobile end-user research.

The context algorithm presented in this paper is pio-

neering because it can be integrated into the end-user

research process than constantly includes thousands of

smartphone users all around the world. This provides both

a huge domain of end-users and at the same time a stan-

dardized process to implement panel studies.

6 Appendix: Other visualizations of the context

algorithm utilization

Figures 8, 9, 10, 11, 12, 13.

Fig. 8 Exemplary output of the

context algorithm


123

Fig. 9 Distribution of active

usage hours

6%

5%

6%

27%

31%

25%

11%

12%

14%

12%

15%

12%

33%

26%

28%

13%

11%

14%

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%

Home

Office

On the move

Percentage of Time

Browsing Messaging Multimedia PIM Voice Other

Fig. 10 Smartphone time

allocation in different contexts

31%

33%

36%

36%

25%

30%

31%

31%

44%

38%

33%

33%

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%

Outbound voice

calling

Inbound voice

calling

Outbound sms

messaging

Inbound sms

messaging



Fig. 11 Person-to-person

service usage in different

contexts


123

References

Abowd GD, Atkeson CG, Hong J, Long S, Kooper R Pinkerton M

(1997) Cyberguide: a mobile context-aware tour guide. ACM

Wireless Netw 5:421–433

Anckar B, D’Incau D (2002) Value creation in mobile commerce:

findings from a consumer survey. J Inf Technol Theory Appl

4(1):43–64

Angehrn A (1997) Designing mature Internet business strategies: the

ICDT model. Eur Manage J 15(4):361–369

ARC Group (1999) Wireless industry showing us the future for the

Internet. In: Wireless Internet: applications, technology and

player strategies (1999–2004)

Balasubramanian S, Peterson RA, Jarvenpaa SL (2002) Exploring the

implications of M-Commerce for markets and marketing. J Acad

Market Sci 30(4):348–361

Barsi T (2002) Disruptive technology vs. disruptive applications.

Telephony Online. http://telephonyonline.com/news/telecom_

disruptive_technology_vs/index.html. Accessed on 6 June 2006

Basaure A (2004) Preliminary research on existing and planned

mobile data service solutions and value systems in leading

markets. LEAD project report. Networking Laboratory, Helsinki

University of Technology

Carmichael DJ, Kay J, Kummerfeld B (2005) Consistent modelling of

users, devices and sensors in a ubiquitous computing environment.

User Model User-Adapt Interact 15(3–4):197–234 (Aug. 2005)

Cerqueira, E, Neto A, Veloso L, Monteiro E, Mendes P (2005) Control

of multi-user services in mobile heterogeneous environments. In:

Proceedings of the third international working conference—

performance modelling and evaluation of heterogeneous net-

works, Het-Nets05, Ilkley, West Yorkshire, UK, July 2005

Chande S (2005) Mobile web services, Department of Computer

Science, University of Helsinki. http://www.cs.helsinki.fi/u/

chande/courses/cs/MWS. Accessed on 20 July 2006

Christensen CM (1997) The innovator’s dilemma. Harvard Business

School Press, Cambridge

Clarke I, Flaherty T (2003) Mobile portals: the development of M-

Commerce gateways. In: Mennecke BE, Strader TJ (eds) Book

mobile commerce: technology, theory, and applications. Idea

Group Publishing, Hershey, pp. 185–201

Dabholkar PA (1996) Consumer evaluations of new technology-based

self-service options: an investigation of alternative models of

service quality. Int J Res Market 13:29–51

Eagle, N (2005) Machine perception and learning of complex social

systems. Doctoral dissertation, Massachusetts Institute of

Technology

WEB Session Duration

5.15

4.024.06

3.48

6.02

4.43

4.96

4.19

5.17

4.13

0.0

1.0

2.0

3.0

4.0

5.0

6.0

7.0

Mean WEB session duration Median WEB session duration

WE

B S

essi

on

Du

rati

on

[M

inu

tes]

Home Office

On the move Home (weekend)

On the move (weekend)

Fig. 12 Average browsing

session durations in different

contexts

32%

25%

35%

27%

28%

21%

42%

48%

43%

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%

Browsing packetdata usage

Email packet datausage

Internet servicespacket data usage



Fig. 13 Internet service usage

in different contexts


123

http://telephonyonline.com/news/telecom_disruptive_technology_vs/index.html

http://telephonyonline.com/news/telecom_disruptive_technology_vs/index.html

http://www.cs.helsinki.fi/u/chande/courses/cs/MWS

http://www.cs.helsinki.fi/u/chande/courses/cs/MWS

Ecosys (2004) First results on the economics of the new service

provisioning in 3G and beyond. The CELTIC initiative, techno-

economics of integrated communication systems and services

FMTC (2004) Mobile services market in Finland 2003 (in Finnish).

Finnish Ministry of Transport and Communications Reports

Giaglis GM, Kourouthanassis P, Tsamakos A (2003) Towards a

classification framework for mobile location services. In:

Mennecke BE, Strader TJ (eds) Book mobile commerce:

technology, theory, and applications, IGI Global, Hershey, pp

67–85, 22 Oct 2002

Heinonen K (2004) Time and location as customer perceived value

drivers. Doctoral Dissertation, Swedish School of Economics

and Business Administration, Helsinki, Finland

Heinonen K, Andersson P (2003) Swedish mobile market: consumer

perceptions of mobile services. Commun Strateg 49(1st quar-

ter):151–171

Heinonen K, Pura M (2006) Classifying mobile services. Presented at

Helsinki Mobility Roundtable, 1–2 June 2006, Helsinki

Holtel S (2006) A timeless way of service classification. Vodafone

R&D white papers. Understand the personality of future

communication services, Vodafone, 29 Mar 2006

Hyvonen K, Repo P (2005) Mobile services in the quotidienne of

Finns (Mobiilipalvelut suomalaisten arjessa), Vox consumpto-

ris—Kuluttajan aani, Kuluttajatutkimuskeskuksen vuosikirja

2005, Savion Kirjapaino Oy, Kerava

Jarvenpaa S, Lang K (2005) Managing the paradoxes of mobile

technology, information systems management. Fall 22(4):7–23

Kaasinen E (2003) User needs for location-aware mobile services.

Pers Ubiquitous Comput 7(1):70–79 (May 2003)

Kaasinen E (2005) User acceptance of mobile services—value, ease

of use, trust and ease of adoption. Doctoral dissertation, VTT

Publications 566, Finland

Korhonen P (2003) Technology leadership. Paper Presented at Nokia

Capital Market Days, New York, 24 November 2003

Kotler P, Armstrong G (1996) Principles of marketing. Pearson

Education, Inc., Upper Saddle River, New Jersey

Li W, Kilander F, Jansson CG (2006) Toward a person-centric

context aware system. In: Proceedings of 1st RSPSI International

Workshop, 7 May 2006, Dublin, Ireland

Meuter ML, Ostrom AL, Roundtree RI, Bitner MJ (2000) Self-service

technologies: understanding customer satisfaction with technol-

ogy-based service encounters. J Market 64(3):50–64

Mitchell K, Whitmore M (2003) In: Mennecke BE, Strader TJ (eds)

Location based services: locating the money. Idea Group

Publishing, Hershey, pp 51–66

Nysveen H, Pedersen PE, Thorbjørnsen H (2005) Intentions to use

mobile services: antecedents and cross-service comparisons.

J Acad Market Sci 33(3):330–346

O’Loughlin, MA (2000) Wireless Internet: opportunities and threats.

Ovum Group

Ojala T, Korhonen J, Aittola M, Ollila M, Koivumaki T, Tahtinen J,

Karjaluoto H (2003) SmartRotuaari—context-aware mobile

multimedia services. In: Proceedings of the 2nd international

conference on mobile and ubiquitous multimedia, Norrkoping,

Sweden

Okazaki S (2005) New perspectives on M-Commerce research.

J Electron Commer Res 6(3):160–164

Pura M, Brush G (2005) Hedonic and utilitarian motivations for

mobile service use. Paper presented at SERVSIG Research

Conference, Singapore, 2–4 June 2005

Raento M, Oulasvirta A, Petit R, Toivonen H (2005) ContextPhone—

a prototyping platform for context-aware mobile applications.

IEEE Pervasive Comput 4(2):51–59

Rogers EM (1962) Diffusion of innovation. Free Press, New York

Smura T (2006) Competition between emerging wireless network

technologies: case HSPA vs. WiMAX in Europe. In: 17th

European Regional ITS Conference, Amsterdam, Netherlands,

22–24 August 2006

SPSS (2002) SPSS 11.5 syntax reference guide, vol 1. SPSS, Chicago

Sullivan MG, Drennan J (2005) Marketing M-services: establishing a

usage benefit typology related to mobile user characteristics.

Database Market Customer Strateg Manage 12(4):327–341

Trevisani E, Vitaletti A (2004) Cell-ID location technique, limits and

benefits: an experimental study. Mobile computing systems and

applications, 2004. WMCSA 2004. Sixth IEEE Workshop on

Volume, Issue, pp 51–60, 2–3 December 2004

UMTS Forum (2001) The UMTS third generation market—phase II:

structuring the service revenue opportunities. Report No. 13.

UMTS Forum, April 2001

Van der Hejden H, Ogertschnig M, van der Gaast L (2005) Effects of

context relevance and perceived risk on user acceptance of

mobile information services. In: Proceedings of the Thirteenth

European conference on information systems (ECIS), Regens-

burg, Germany

Vanjoki A (2003) Multimedia—seed for growth. Nokia Capital

Market Days 2003. http://media.corporate-ir.net/media_files/

webcast/2003/nov03/Nokia/PPT/Drivingconsumermultimedia.pdf.

Accessed on 20 July 2006

Velez FJ, Correia LM (2002) Mobile broadband services: classifica-

tion, characterisation and deployment scenarios. IEEE Comm

Mag 40(4):142–150

Verkasalo H (2005) Handset-based monitoring of mobile customer

behavior. Master’s Thesis Series, Networking Laboratory,

Department of Electrical and Telecommunications Engineering,

Helsinki University of Technology, Espoo, Finland, September

2005

Verkasalo H (2007a) A cross-country comparison of mobile service

and handset usage. Licentiate’s Thesis, Helsinki University of

Technology, Networking Laboratory, February 2007

Verkasalo H (2007b) A snapshot on the evolution of Finnish mobile

service market. In: Conference on telecommunication techno-

economics (CTTE) 2007, Helsinki, Finland, 14–15 June 2007

(Submitted)

Verkasalo H, Hammainen H (2006) Handset-based monitoring of

mobile subscribers. Presented at Mobility Roundtable, Helsinki,

Finland, 1–2 June 2006

Verkasalo H, Hammainen H (2007) A handset-based platform for

measuring mobile service usage. INFO: J Policy Regul Strateg

9(1):80–96

Verkasalo H, Smura T, Kivi A (2007) A framework for analyzing the

usage of mobile services. Networking Laboratory, Helsinki

University of Technology (Submitted)

Vesa J (2005) Mobile services in the networked economy. IRM Press,

Hershey

Williamson J (2003) Content or brand? Paper presented at the 3GSM

World Conference, Cannes, France

Yoon JI, Suh HS, Leem CS (2003) A study on mobile business

customer satisfaction. Korea Institute for Industrial Engineering,

April 2003


123

http://media.corporate-ir.net/media_files/webcast/2003/nov03/Nokia/PPT/Drivingconsumermultimedia.pdf

http://media.corporate-ir.net/media_files/webcast/2003/nov03/Nokia/PPT/Drivingconsumermultimedia.pdf

Contextual patterns in mobile service...

Documents

Transcript of Contextual patterns in mobile service...