A User-Centered Design Approach to Self-Service Ticket Vending Machines

138 IEEE TRANSACTIONS ON PROFESSIONAL COMMUNICATION, VOL. 56, NO. 2, JUNE 2013

Case Study

A User-Centered Design Approach to Self-Service TicketVending Machines

—KARIN SIEBENHANDL, GÜNTHER SCHREDER, MICHAEL SMUC, EVA MAYR, AND MANUEL NAGL

Abstract—Since their introduction, self-service ticket vending machines (TVMs) have become an increasinglyimportant distribution channel in the public transport sector, progressively replacing the traditional ticket counter. Ina public transport setting, where ticket counter closures have left different groups of people dependent on TVM tomeet their mobility needs, a single, effective system is required. Research questions: (1) Which barriers do currentlyhinder the usage of TVM? (2) Which requirements should a barrier-free TVM fulfill? (3) How can we design a newself-service TVM for a nationwide public railway company? (4) How can we ensure that the usability and userexperience (UX) is high for all users, especially for those with low levels of technological affinity? Situating thecase: Most other studies on the use and usability of TVMs were conducted as post-hoc evaluations. In contrast,our case study presents a user-centered design (UCD) approach that takes the needs of the different target groupsinto account throughout the whole development process. Theories and concepts that guided the case included UCD,which involves alternating test and evaluation loops that actively involve users to create a usable product and UX,which describes the quality of the experience a person has when interacting with a specific computer system using aspecific interaction technique. Methodology: More than 250 participants were involved in focus groups, observations,interviews, and experiments from the very first stages of development. Interface designs were presented to thefuture end users to obtain their feedback, with the results fed back into the design process. About the case: Aprototype for a novel generation of TVM was developed in three phases: First, the context of use was analyzed.In the second phase, we conducted a requirements analysis. Third, different hardware and software interactiondesigns were iteratively tested and evaluated. The resulting prototype met the requirements of most user groups,though further adjustments are necessary. Conclusions: The UCD approach proved to be a valuable framework forthe development and design of self-service systems.

Index Terms—Self-service technologies, senior user, ticket vending machines (TVMs), user-centered design (UCD),user experience (UX).

INTRODUCTION

There is an observable trend across Europe tocontinuously reduce the number and openinghours of serviced ticketing for public transport.This forces passengers to use self-service ticketingchannels like the internet, mobile ticketing, or—asin the case of this study—ticket vending machines(TVMs).

These self-service channels pose a barrier—espe-cially to older passengers [1]–[4]. Why is this thecase? The current generation of ticket machinesin Europe (often developed around 15 years ago)mainly displays “machine-generated logic” at theinterface to the customer. No focus was placedon user friendliness and the user experience (UX)

Manuscript received June 13, 2012; revised March 14, 2013;accepted March 25, 2013. Date of publication May 13, 2013;date of current version May 20, 2013.The authors are with the Center for Cognition, Informationand Management, Danube University Krems, Krems 3500,Austria (email: [email protected]; [email protected]; [email protected];[email protected]; [email protected]).Color versions of one or more of the figures in this paper areavailable online at http://ieeexplore.ieee.org.

IEEE 10.1109/TPC.2013.2257213

during their development. As a consequence, peoplewith low technological affinity have difficulties inusing these machines. Given that transport policiesseek to encourage the increased use of publictransport by the broad population, reducing theseaccess barriers will be one of the major tasks facingtechnical communicators when developing suchsystems.

This paper presents a case study on thedevelopment of a prototype for a new generationof user-friendly self-service TVMs, the INNOMATproject. This prototype will be the basis forthe industrial development of the new TVM forthe Austrian Federal Railway [ÖsterreichischeBundesbahnen (ÖBB)]. The project itself serves as ashowcase for a user-centered design (UCD) processfor self-service systems. When service providersdevelop a self-service system intended for use by thebroad public, users should be included throughoutthe entire development process—right from itsearly beginning. The aim of such a developmentprocess should not only be to develop a systemwhich is usable by most users, but to develop asystem which leaves all passengers with a positivefeeling after use. Accordingly, the INNOMAT project

0361-1434/$31.00 © 2013 IEEE

SIEBENHANDL et al.: UCD APPROACH TO SELF-SERVICE TVMS 139

focused not only on usability, but also on the UXof different user groups.

The research questions of the case study were:

• Which barriers do currently hinder the usage ofTVM?

• Which requirements should a barrier-free TVMfulfill?

• How can we design a new self-service TVM for anationwide public railway company?

• How can we ensure that the usability and UX ishigh for all users, especially for those with lowlevels of technological affinity?

In this paper, we situate our case study in thecontext of other research on barriers in self-serviceticketing and discuss relevant theoretical concepts,namely, UCD and UX. We then describe how thiscase was studied and how we applied the UCDprocess. Next, we present our case, outlining eachstep of the UCD process with its main results. Inthe final conclusions section, we discuss how theUCD approach helped us to reach our aims andexamine those factors which proved to promote orhinder the development of a novel, self-service TVM.

SITUATING THE CASE

This section situates our case study. First, weoutline the barriers to the use of self-service TVMalready identified in prior case studies. We thendescribe UCD in its context as a methodology whichallows user needs to be addressed during the entiredesign process. Finally, we discuss UX as a factorwhich can promote a positive experience with TVMand lead to their repeated use.

How Literature Was Selected TVMs have notbeen a subject of major research interest untilnow. We therefore conducted a broader literaturereview and searched not only for prior research onour specific target technology (TVM), but also onother self-service technologies—for public transportor other services. In doing so, we tried to identifyexisting barriers to the use of TVM and successfactors in existing design solutions and effectivedevelopment processes.

Self-service technologies and UCD cannot bereduced to one single discipline. Likewise, aliterature review of this topic cannot be reducedto one single database. We therefore conducted abroad literature search in databases on computersciences, engineering, and social sciences, but also

a more open search on cross-disciplinary databaseslike Google Scholar.

Barriers to the Use of Self-Service TVMs Manysubgroups in today’s society are not skilled atusing new technologies. Based on their negativeexperiences with technology in everyday life, theyoften develop a tendency to avoid new technologies.The resulting lack of technological expertise canbuild barriers that affect mobility and lifestyle.This certainly holds true for the elderly and othersocioeconomic groups disadvantaged by the digitaldivide, since they rely, in particular, on publictransport [1].

Prior negative experiences with everydaytechnologies represent one of the strongest barriersto the use of self-service TVM. Research into ticketqueues at 12 major railway stations in Great Britainsupports this assertion. Passengers who couldhave bought their tickets from a TVM confirmedthat their decision to purchase at the ticket counterwas driven by a lack of confidence in their abilityto use the machine as well as a lack of confidencein their “ability to select a ticket at the appropriateprice with the necessary validity or relevant route”[2, p. 4].

Subasi et al. [3] come to a similar conclusion intheir study of usage barriers and perceptions ofan online ticketing service for a nationwide publicrailway company in Austria. With regard to theoptimization of an online system, they concludethat:

it is necessary to develop a system which is notonly universally accessible, but also satisfiesthe specific expectations of senior users interms of usage patterns and their specific needsas well as different perception models.

To overcome the lack of confidence among olderpassengers, the ALISA project [4] added trainingmodules to a TVM. The training helped userslearn how to use the machine and increased theirtechnological self-efficacy—independent of age.

These case studies on self-service TVM wereconducted as post-hoc evaluations [2]–[4]. None ofthem goes through the entire design process—a gapwhich our case study can close by using a UCDapproach. The main benefit of such an approachis that the design can already be aligned with theneeds of many different user groups at an earlystage, removing the need for post-hoc adaptationsas in the ALISA project [4].

140 IEEE TRANSACTIONS ON PROFESSIONAL COMMUNICATION, VOL. 56, NO. 2, JUNE 2013

UCD Approach In the UCD approach, thedevelopment of the design process is plannedin alternating test and evaluating loops whichactively involve users [5, p. 588]. More specifically(and according to the ISO 13407 standard onhuman-centered design), this includes five essentialsteps:

• Plan the human-centered design process• Understand and specify the context of use• Specify the user and organizational requirements• Produce design and prototypes• Carry out user-based assessment includingformative evaluation

To ensure that a self-service system like a TVMmeets the demands of a system intended for use bythe general public (a broad target group), the UCDprinciples [5] should be adapted as follows:

• Interdisciplinarity: The conceptual constructionand design of a self-service system as well asthe functional aspects of the user interfaceshould be developed by an interdisciplinaryteam of engineers and specialists forergonomics and accessibility, media design, andhardware/software development.

• User-oriented design: To guarantee the designof a user-oriented system (both hardware andsoftware), the target groups should be involvedfrom the very first stages of development. Theneeds and requirements of different target groupsshould already be taken into account during thedevelopment stage; for example, to accommodateexperienced users who want fast processing withfew steps as well as inexperienced users whoneed comprehensive, intuitive prompting viaaudio-visual information channels.

• Iteration of design solutions: Design approachesand solutions should be presented to thefuture end users to generate feedback and theresults fed back into the design process. Userstudies should systematically involve users fromthe beginning to the end of the project—frompaper mockups of initial designs, through earlyclickable prototypes to a working prototype of anactual TVM.

• Allocation of function between the user and thesystem: A self-service system for the broad publicshould adapt to the user wherever possible. Anyinput required by the user should be kept tothe necessary minimum. At the same time, thesystem has to provide the user with as muchfeedback as necessary to ensure he/she is awareof the selections he/she has made.

By translating the UCD principles to (thedevelopment of) self-service TVM, theseassumptions all essentially relate to the qualityof use of such machines. How quickly can I geta ticket? How complicated is the process? Howconfident am I that I will be able to get the bestvalue ticket from this machine and pay the correctprice? To what extent does the system’s purchaseprocess correspond to the mental models thedifferent target groups have of such a process? Towhat extent can the design of the TVM supporta purchase process with positive connotations?These questions, in turn, emphasize the importanceof research into UX for self-service technologysystems.

User Experience (UX) UX describes the qualityof the experience a person has when interactingwith a specific computer system using a specificinteraction technique [6]. According to Hassenzahland Tractinsky, UX incorporates different aspectslike social experience, ambient conditions, joy ofuse and aesthetics, as well as factors like perceivedprivacy, security, and trust:

UX is a consequence of a user’s internalstate (predispositions, expectations, needs,motivation, mood, etc ), the characteristicsof the designed system (such as complexity,purpose, usability, and functionality) and thecontext (or environment) within which theinteraction occurs. [7, p. 95]

In the TVM context, factors like social experienceand ambient conditions would seem to beparticularly important. In this respect, technologyaversion is caused or intensified at least to a certainextent by negative social experiences. It is notsurprising that elderly users, in particular, oftenfind themselves confronted during the purchaseprocess with impatient reactions and socialpressure from the other people waiting in line.

In general, there is an increasing demand forpublic self-service systems like TVM (such ase-government or e-health systems), and theirprevalence will increase in the future. In thefollowing sections, we present a multifaceted casestudy that describes how a new generation of TVMwas designed to best meet the needs of differentuser groups. The findings of this study can serve asindicators for future research and design questionson the relations between self-service technologiesand UX.

SIEBENHANDL et al.: UCD APPROACH TO SELF-SERVICE TVMS 141

Fig. 1. Applied UCD process.

HOW THIS CASE WAS STUDIEDIn this section, we present the methodology usedin our case study: First, we discuss our choiceof research methodology. We then describe theparticipants in the case study, explain how datawere collected and analyzed in the UCD process,and demonstrate how we ensured credibility andtrustworthiness.

Choice of Research Methodology Fig. 1 providesa general description of our UCD approach, theindividual phases and steps in the project, and themethods used.1 As can be seen, we used a wideset of different methodologies in the course of theproject to best inform the next design phases andanswer the relevant research questions in eachphase.

In the first project phase (Requirements), ourresearch focused on (1) looking into existingbarriers to the usage of TVM, and (2) identifyingrequirements for a barrier-free TVM. To identify asmany existing barriers as possible (CoU), we decidednot to restrict ourselves to one single methodology(a procedure often necessary in economic contexts,when resources are more limited), but insteadused different methodologies with different aims.Accordingly, we reviewed existing literature on theuse of technology by people with special needs

1A detailed description of each phase, including itsmethodology, will be given in the section “About the Case.”

Fig. 2. Project team.

(vision impaired, mobility impaired, the elderly) andtheir requirements for using a TVM. In addition, weinvited experts and stakeholders to participate in afocus group to add their experiences to the materialgathered. To identify interaction problems andbarriers to the use of the existing TVM in Austria,we conducted observations (in situ, video analysis,logfile analysis) and interviews at railway stations.To gather ideas on how to overcome these existingbarriers, we carried out a competitor analysis ofother TVM and self-service technologies througha literature survey and physical observations atrailway stations across Europe. Regular meetingswith the project consortium were used to collate allthe information collected into a list of interface andhardware requirements [RA].

142 IEEE TRANSACTIONS ON PROFESSIONAL COMMUNICATION, VOL. 56, NO. 2, JUNE 2013

Fig. 3. Distribution of age and technological affinity among participants. Cutoff scores depicted as lines separatethe participants into four experimental groups.

In the second project phase (Design andEvaluation), the research focus lay on (3) designinga new self-service TVM system for the ÖBB, and(4) ensuring that the usability and UX of this newTVM was high for all users, especially for those withlow levels of technological affinity. After definingthe use cases, we conducted brainstorming andcreativity sessions to design first wireframe layouts[D1] and gained user feedback on these designsat a very early stage in the design process fromfocus groups and experts [P]. Although we were notable to conduct an experiment or any observationsat this stage, this early feedback on the designsallowed us to avoid cost-intensive developmentactivities in the wrong direction. In later stages, wetested three different software mockups [D2] in anexperiment [E1] and developed the most usableof these mockups for inclusion in an integratedhardware and software prototype [D3] for the lastround of user testing [E2].

Participants Two key principles of UCD are (1)involving the users throughout the whole process,and (2) working in an interdisciplinary team.Accordingly, a number of different people wereinvolved in this design process. The regular projectteam in the INNOMAT consortium was made up ofan interdisciplinary group of hardware and softwaredevelopers, media designers, usability experts, and

representatives of the ÖBB. (See Fig. 2.) During thedifferent phases, we called in external experts andstakeholders to complement the know-how of theproject team, for example, on UX, accessibility, andthe needs of different target groups.

For user testing, we strived for two differentcharacteristics in our participant sample: (1)diversity in age, and (2) diversity in technologicalself-efficacy (SE). Since public transport passengersrange in age from the very young to the very old,we also strived to include a wide age range in ourparticipants (13 to 84 years, see Fig. 3). We sought,in particular, to ensure that half of our participantsfell into the older age bracket (above 55 years), thusincluding those people who had been identified ashaving problems with TVM in the past. Anothergroup that had been identified as having problems(though often overlapping with the elderly group)was people with low technological affinity. Wetherefore sought to include participants who scoredhigh as well as participants who scored low on ourtechnological affinity scale.

We used different strategies to recruit testparticipants: In the railway station setting[observations and interviews (CoU)], we used a timesampling method: We observed TVM users on fouroccasions (weekday–weekend, urban–rural stations)

SIEBENHANDL et al.: UCD APPROACH TO SELF-SERVICE TVMS 143

and categorized them with respect to their age andhow they approached the TVM (cautious—targetedfast interaction).

For the studies in our usability laboratory,we invited local participants using differentstrategies: First, we used our own database of testparticipants, which turned out to contain mainlyyoung people with high technological affinity. Tocombat this bias, we used alternative strategies likenewspaper adverts and requests to mailing listsannouncements. We also contacted local retirementhomes. Each potential participant answered a shortscreening questionnaire on technological affinity:Only those with either high or low technologicalaffinity were included in the studies. Some peopleparticipated in both experiments 1 and 2, but thisnumber was restricted to 45%. Participants receiveda payment of 25 Euros to cover their expenses.

Ethics approval was required neither by our owninstitution nor by our funding agency.

How Data Were Collected Wherever possible,we collected data in a standardized manner toallow us to analyze them efficiently and conductstatistical analysis to produce a generalizationof results. To gain a more complete picture, wecollected qualitative data that were not included inthe structured recording protocols.

• We conducted pretests and created observationprotocols from the behavioral patterns observedfor any observations carried out. Additionalbehavioral patterns observed during the maintests were recorded in a free format. Similarly,we created structured protocols of features ofinterest for the competitor analysis. To illustratethese protocols, we took pictures of the differentTVM and their features of interest.

• We created structured guidelines and protocolsheets to use in the interviews. We also recordedthe interviews to allow us to transcribe importantsequences and amend the protocols if necessary.

• Focus groups were conducted in the usabilitylaboratory and video recorded with twodome cameras and audio recorded with twomicrophones. Two members of our researchteam led the group in line with a predefinedstructure. Another two observed the discussionin an adjacent room, took down notes, anddocumented major results.

• For experiments 1 and 2, participants wererandomized to the different experimentalconditions. Experiments were conducted in theusability laboratory and lasted between one and

two hours. Participant interaction with the TVM(mockup) was logged, behavior was observed,and think-aloud-protocols were recordeddigitally. Two members of the experiment teamwere present to instruct the participants andnote down any observations and importantstatements.

• Logfile analysis from the train station and fromexperiment 2 was conducted by the softwarepartner in the project.

How Data Were Analyzed We input quantitativedata into Excel and SPSS and analyzed these datastatistically. In essence, we computed frequenciesto gain an overview of the distribution of features,interaction problems, and successful interactions.To generalize our findings, we computed statisticaltests of mean comparisons (ANOVA) betweenparticipants in the elderly and young age groupsand those with high and low technological affinity.

We conducted content analyses of qualitative data.We bundled the answers to specific questions andsearched for common topics as well as commonbehavioral patterns or features.

Ensuring Credibility and Trustworthiness Toensure high reliability, we strived to collect andanalyze data in an objective way. First, wherepossible, we structured data collection and usedstandardized protocols (see section “How datawas collected”). Second, we recorded data wherepossible to allow us to review the ratings ofobservation, focus groups, and interview protocols.In those cases where data recording was notpossible, we used two independent observers (thedirect versus cautious approach to the old TVM wasrated, for example, by two independent observers atthe railway station, yielding an agreement of morethan 90%). Third, we conducted an internal trainingsession for the experiment team, observers, and forthe coding team to ensure that instructions werealways given in a similar manner, that protocolswere understood in a similar way, and that datawere coded in the same way. Fourth, we selectedand negotiated all methodological decisions in ourgroup of four to five usability experts to look forpotential pitfalls and problems.

ABOUT THE CASE

In the following section, we will provide astep-by-step account of how we proceeded in theINNOMAT project to develop a prototype for a novel,barrier-free TVM for the ÖBB. In doing so, we will

144 IEEE TRANSACTIONS ON PROFESSIONAL COMMUNICATION, VOL. 56, NO. 2, JUNE 2013

Fig. 4. Final design of the new TVM interface (left) and hardware (right).

present the problem, our solution, and how wedeveloped it.

Problem Current TVMs do not meet therequirements of many of their users. In the publictransport context, they tend to be aligned withbusiness processes and mirror the thinking of ITspecialists rather than the thinking of travelers.In the past, users were not involved in thedevelopment process for TVM. User studies wereonly conducted after the design was completed,and major redesigns were not possible at such alate stage.

To involve the users from the very start of thedevelopment process, this case study was not anindustrial development, but was instead proposedand funded as a research project—with all of theassociated constraints that come with such aproject. In our case, the funding was cut, and wewere therefore not able to conduct a final evaluationin situ in a railway station setting.

Solution In an iterative process, we developeda novel barrier-free TVM using a UCD approach.

The TVM was aligned with the requirements andthe thinking of the broad public who use suchmachines, especially the elderly and people withlow technological self-efficacy.

The main screen of the resulting TVM only hasa small number of buttons, which are clusteredusing different colors for different functions [Fig. 4,(top left)]. A “search engine-like” destination inputbar is the most salient element on the screen. Thepurchase process interface narratively maps a trainjourney and by resembling the cognitive script oftravelling is intuitive for many people and manydifferent user groups [Fig. 4, (bottom left)]. Wherepossible, the TVM hardware reacts to user input ina smart way: It automatically adjusts to the user’sheight, activates relevant areas, and uses lightingto direct the user to these areas [Fig. 4, (right)].Novel features have been included if they makethe purchase process easier, but no cutting-edgetechnologies have been used that are of no benefitto the users.

SIEBENHANDL et al.: UCD APPROACH TO SELF-SERVICE TVMS 145

Fig. 5. INNOMAT project facts about the solution.

Fig. 5 provides an overview of the key facts andfigures regarding the INNOMAT project.

Process for Developing the Solution The processfor developing the solution involved many phases.Fig. 1 identified the key phases in this process,which we followed when developing this project. Thefollowing sections describe our process in detail.

CONTEXT OF USE (COU)

All good design processes begin by examining thecontext of use, and the INNOMAT project was nodifferent. In our case, this examination had tofocus on the current situation at railway stations inAustria, the way people used the existing systemsavailable at these stations, and their attitudetowards these machines. We also aimed to identifyother good and bad examples of TVM in Europeand to determine whether any other work had beencarried out elsewhere to design a user interface thatbetter accommodates the broad needs of the targetgroups (in particular, nontechnical railway users,persons with reduced mobility). In order to shedlight on these issues, we used a variety of methodsto ensure we obtained a holistic evaluation of thestatus quo.

Literature Review: In a first step, we collectedthe special needs and requirements of differenttarget groups already identified in prior research.The findings of this literature review establishedlinks between the context of use (the railwaystation) and the users of the public transportnetwork. From existing research dealing withthe particular needs of the disabled, we wereable to identify and incorporate a number ofspecific design requirements that have to be takeninto consideration to allow people with visionimpairments or wheelchair users to actually use

(a new generation of barrier-free) TVM in the firstplace. In addition, our literature review includedpreparatory work to accommodate the specialneeds of the elderly, which frequently result fromthe changes to their cognitive abilities experiencedby people in this age group. A detailed descriptionof the findings of this phase can be found in [8].

Observations and Interviews: In order to specifyand understand the target users, we conducted afield study in which we observed 50 people at tworailway stations and interviewed an additional 50people: 25 as they used TVM and the other 25 infront of the ticket counter. Our aim here was toidentify user groups who had problems using themachines and determine where these problems lay.To assess their experiences, we asked these peopleabout their levels of satisfaction with and reasonsfor choosing or avoiding a TVM.

Our observations and interviews showed nooccurrence of “typical accessibility or usabilityproblems” like font size, contrast, or button size,even among elderly users. In contrast, most peoplewere able to use the TVM without encounteringsuch problems. However, to ensure an all-roundpositive UX, we felt it was important to focus on theproblems encountered by those people who did notsucceed in using the machines.

The observations and interviews revealed a numberof serious barriers to the use of TVM, above allamong older and middle-aged passengers. Theyoften approached the TVM with great caution andsometimes had to cancel their purchase process.Some customers in these age groups, in particular,had little confidence in their ability to successfullybuy a ticket at a self-service terminal. When askedwhy they avoided the machines, they referredto bad experiences, doubt in their own abilities,and mistrust in the technology. (For more details,

146 IEEE TRANSACTIONS ON PROFESSIONAL COMMUNICATION, VOL. 56, NO. 2, JUNE 2013

see [9].) They lack positive experiences of usingeveryday technologies.

Social Cognitive Theory offers an explanation forthe underlying emotional and motivational factorresponsible for avoiding everyday technologies:Self-efficacy (SE) is the belief “in one’s capabilitiesto organize and execute the courses of actionrequired to produce given attainments” [10, p. 3].In Czaja et al. [11], computer SE was an importantpredictor for the use of technical devices, and wasmediated by computer anxiety. Czaja et al. (ibid.)noted that people with a low level of SE had a lowerprobability of using a technology. Since seniorcitizens are frequently represented in this group,the authors stressed the importance of usingtechnology that allows senior citizens to experiencesuccess and build up their confidence in their ownabilities. Consequently, low levels of technologicalliteracy and the underlying emotional factor ofavoiding the terminals—low levels of computerSE—were subsequently taken into consideration inthe creation of the new interface.

Video Analysis: Analyzing anonymous videorecordings of ticket purchase scenarios at alarge railway station in Vienna provided us withfurther important insights into the location ofTVM, the distance between machines, the layoutof hardware elements (such as payment elementsand dispensing drawers), as well as the behaviorof customers during the purchase process. In theanalysis of these videos, particular focus was givento the general behavior exhibited by customerswhen standing in line or selecting a machine: Ingeneral, they tried to maintain a social distance toother customers by selecting the machine that wasfarthest away from other customers. When a groupof customers used one TVM, the adjacent machinewas also blocked due to the narrow layout of themachines. Another problem observed was thatpeople had to bend down to retrieve their ticketsfrom the dispenser. This was a particular problemfor elderly and mobility-impaired customers.

Log File Analysis: Purchase process logs forÖBB TVM at the Vienna South railway station(Südbahnhof) were used to analyze the currentsystem and current configuration of the purchaseprocess. Logs of 144 purchase processes (onOctober 14, 2008 from 06:24 to 10:56 am) wereanalyzed. Of these, 61 were cancelled prior tocompletion of the actual purchase, with 54% ofprocesses cancelled on the initial start screen and20% on the selection screen. The start screen

displays around 25 buttons, which seems tooverwhelm customers in their search for the rightproduct. The selection screen also required thatticket specifications be input in a predefined orderthat was not transparent to the customers. Theseobservations led us to the conclusion that a greatdeal of emphasis had to be placed on the redesignof the start and selection screens.

Competitor Analysis: Important information oncurrent state-of-the-art TVM was also obtained fromour broad competitor analysis. The analysis datawere gathered between December 2008 and April2009 at railway stations in Austria (such as Vienna,Linz, and Bregenz) and nine other European cities(Munich, Frankfurt, Zurich, Brussels, Amsterdam,Strasbourg, Lille, Mikkeli, and Barcelona), andthe US (New York and Boston). These locationswere selected based on preliminary research andrecommendations. The site analysis focused onthe location/position of TVM in stations (easeof identification, accessibility factors), hardwaredesign (such as height adjustability and options foruse by the visually impaired), and user interfacedesign (such as purchase process and accessibilityfactors). The data were supplemented by materialprovided on request by the respective transportoperators or published on the internet.

One particularly interesting TVM was identified inthe Netherlands. It displays the entire purchaseprocess on one screen, thereby providing thecustomer with a good overview. It also providesdirect visual feedback on any selections made.Interestingly, the development of this TVM wasaccompanied by user tests and aligned with gestaltprinciples [12].

Accessibility Workshop: A special accessibilityworkshop was held on March 23, 2009 atthe Danube University Krems in Austria. Thisworkshop was attended by accessibility experts,representatives of the various target groups (suchas organizations for the visually impaired ororganizations for people with mobility restrictions),and representatives of the ÖBB in its capacity bothas project leader and largest current deployer ofself-service TVM in Austria. Its aim was to establishthe specific requirements, needs, and context of useof the different target groups. In the course of theworkshop, the TVM’s importance as a sales channelfor different target groups was discussed along withthe major hardware and software design problemsand possible solutions. In addition, the differentsystems identified in the competitor analysis wereassessed.

SIEBENHANDL et al.: UCD APPROACH TO SELF-SERVICE TVMS 147

An important outcome that stemmed from thisworkshop was the decision not to focus ondeveloping a TVM for blind users: This user groupprefers other sales channels where they can usetheir personal assistant technologies. Thus, thecosts of developing an interface for the blind wouldbe too high in relation to the benefits.

Regular Meetings: During regular project meetings,the results of the analyses to specify the contextof use were discussed within the project teamand implications derived for the requirementsspecification. At these meetings, the project teamworked not only on developing its ideas andvisions, but also on issues of technical feasibility,maintenance, and organizational implications.

REQUIREMENTS ANALYSES (RA)

The results of the first phase (context of use)were condensated into a list of requirements fora barrier-free TVM. Specific tasks of this phaseincluded identifying interface and hardwarerequirements.

Interface Requirements: As far as the interfacedesign was concerned, it quickly became evidentthat the biggest challenge would lie in reducingthe complexity (such as establishing a purchaseprocess that would be understandable to all userswithout requiring high computer literacy andknowledge of fare structures).

Some of the requirements encountered actuallycontradicted one another [8]. While users with littletechnological literacy argued for a step-by-stepsystem that guides users through the wholeprocess, expert users (such as commuters whoarrive shortly before their train is due to depart andwant to buy their tickets quickly) also have to betaken into consideration. Potential solutions to thisdilemma could include a personalized system (suchas offering cardholders the opportunity to simplyinsert their cards and buy frequently purchasedtickets in a single step).

In the workshops and interviews with targetgroup representatives, the question arose as towhether the TVM actually must/could meet allof the requirements or whether some of thesecould be better provided and served using othersales channels. People with mobility restrictionsfrequently require assistance to board a train and,therefore, generally buy their tickets at the ticketcounter, since they have to request this assistanceat the station in any case. However, ticket counters

are the only alternative sales channel available tonontechnical customers (in our case, primarilysenior citizens). As a result, a readily accessible,easy-to-use interface is important for this targetgroup.

Hardware Requirements: The way the terminalsare embedded into the station infrastructure is afurther important point that became particularlyapparent in the best practice analysis. The keyelements that need to be considered here arethe position and accessibility of the TVM inthe station and the ability to purchase a ticketwithout impediment, such as by enabling users tomaintain a proper social distance and preventingglare. Existing accessibility guidelines providedinformation on positioning machines to provideadequate clearance and accessibility for wheelchairusers and emphasized the need to mark or indicatethese features.

While the layout of hardware elements is, of course,affected by the technical framework and applicablenorms, the layout itself also has a significantinfluence on usability. Our video analysis showedthat the dispensing drawer on the current ÖBBTVM was very low, and that a large number ofpeople had to bend down to get their tickets. Otherrequirements, which have received surprisinglylittle attention to date, relate to the context of use,such as facilities for people with luggage, prams,and pushchairs. The use of clear symbols to identifyspecific hardware elements is a further importanthardware and layout aspect, in particular, whenit comes to tailoring the system to the needs ofvisually impaired users.

While current self-service terminals often have asmaller sibling to allow wheelchair access, thisrather expensive strategy fails to account for thevarious heights needed by users and is not afinancially viable solution for general use at allrailway stations in the future. In UCD processterms, the requirements analyses (RA) correspondsto the step “Specify the user and organizationalrequirements.” In the INNOMAT project, thespecific requirements of the individual targetgroups were assessed for technical feasibility andmaintainability. From a software design perspective,the results of this assessment showed that thepurchase process had to be made as flexible aspossible. From a hardware perspective, the resultsof the study also produced recommendations forsystem maintenance and addressed other aspectslike burglar-proof construction, cleaning options,and other maintenance issues.

148 IEEE TRANSACTIONS ON PROFESSIONAL COMMUNICATION, VOL. 56, NO. 2, JUNE 2013

The results of the RA, together with the analyses ofthe context of use (TVM in railway stations) (CoU)and the different user groups, were used to createa detailed catalog of requirements. In addition toaccessibility and usability factors, this catalog alsocontained some UX aspects, relating, in particular,to the issue of support for the different targetgroups.

DESIGN AND EVALUATION

The next steps in the project were UCD stepsfour and five (“Produce design and prototypes”and “Carry out user-based assessment”). Thegoal here was to identify and develop newapproaches and designs for the TVM’s hardwareand software. The methods employed—from theinitial creativity workshops and wireframes layoutsin the early stages of the project, through to theclickable prototypes and subsequent integratedprototype—were all aimed at systematicallyinvolving the users. Solutions were presented to,tried out by and discussed with the future targetgroups, with the results fed back into the designprocess.

Definition of Use Cases: To structure andsubsequently test the purchase process, fouruse cases were defined in cooperation withrepresentatives of the ÖBB. While these cases wereall based on current ÖBB ticket sales statistics,care was also taken to include a more complexpurchase scenario that would not have beenpossible with the current system—tickets for asmall group of individuals travelling together.Emphasis was also placed on ensuring that theusers would receive a choice of different offers tosimulate the somewhat complex fare structure andthe corresponding selection by the user.

The four selected use case scenarios were asfollows:

(1) Standard ticket for a single person.(2) Discount ticket for a family of two adults and

one child.(3) Ticket for a single person for a return journey

in the local transport network.(4) Tickets for a group of three adults—including

one travel card holder—for a return journey inthe local transport network.

Design 1 [D1]: By the end of Design Phase 1,several mockups had been developed based onideas developed at workshops attended by membersof the target groups and media/usability experts.

Brainstorming and Creativity Workshop: In aworkshop attended by eight media and usabilityexperts from the Danube University Krems, a rangeof creative ideas for a new TVM was developed andcollected. These included new ways of selecting adestination, such as by entering the target stationin a search bar on the start screen (similar to theGoogle Toolbar) or selecting it on a map. The TVMwould then display a selectable list of possibletrain connections. Another suggestion was an“intelligent” TVM which recognizes a customer,instantly provides a list of his/her most recentjourneys, knows his/her current position, andoffers the best deals to the customer’s mostfrequent destinations from the current location.

Wireframes Layout: Based on the ideas collectedand the catalog of requirements, five alternatives fora new start screen were developed in cooperationwith a group of media designers. Fig. 6 showsa wireframe of the start screen used in thecurrent ticket systems (left, frequently criticized asoverloaded) alongside one of the new “less cluttered”alternatives (right, which also incorporates some ofthe features discussed at the creativity workshop).

Pilot [P]: In a next step, we organized aneight-person focus group with members of thespecific target groups (four senior citizens and fouryounger people, two members of each group withlittle technological literacy, see Fig. 3). We alsoorganized a second focus group with eight usabilityexperts. The primary aim of the focus groups wasto gather feedback on the wireframes layout ofthe start screens and obtain ideas and feedbackregarding the structure of the purchase process.Special focus was placed in the discussions onhow the system could enhance UX. What do peoplethink about when they approach a TVM? Howcould the start screen convey the message thatthe system is easy to use? The key responseshere were that the start screen should be clearlystructured and that it should not be overloaded orconfusing. The first focus group also voiced certainmisgivings about the map-based selection of thedestination (“You have to select the destination onthe map? I’m not very good at geography.”). Thesearch bar was received in a particularly positivemanner and described in general—even by thosepeople who virtually never used a computer—as arecognizable and easy-to-use function. We thereforeimplemented the search bar in the subsequentprototypes.

Design 2 [D2]: In Design Phase 2, prototypes werecreated by the design partner and the software

SIEBENHANDL et al.: UCD APPROACH TO SELF-SERVICE TVMS 149

Fig. 6. Wireframes of the start screens of the current TVM (left) and one of the new designs (right).

Fig. 7. Prototype A—overview (left) and passenger selection submenu (right) screens.

partner in the project team. Following the focusgroups, where the wireframes had been discussedat length, design workshops were subsequentlyarranged within the project team (media designers,human–computer interaction (HCI) experts, ÖBB).These workshops were very productive, with anexchange of ideas that led to the creation of threedifferent prototypes. The following could be selectedon each of these four prototypes: destination,different types of passengers (adults, children,families, dogs ), date of travel, and ticket class.

Prototype A—“Virtual Ticket”: The first interfacewas a contemporary adaption of the TVM systemalready in place at ÖBB stations. It featured aticket-like overview and typical human–computerinteraction elements like submenus and drop-downmenus. (See Fig. 7.)

Prototype B—“Netherlands Clone”: This system wasa modification of the easy-to-use system currently

used in the Netherlands. All possible options canbe seen immediately and the current selection ishighlighted. To account for the higher complexityof the Austrian fare system in comparison to itscounterpart in the Netherlands, we developed anovel person configuration approach in whichpassengers could be selected graphically. (SeeFig. 8.)

Prototype C—“Train Metaphor”: Although largelysimilar to prototype B, this option extended thegraphical metaphor by using a train narrative andanimations. The selected passengers “move” into atrain, and any options selected are visualized bytrain wagons moving into the correct position. (SeeFig. 9.) More detailed information on this prototypecan be found (and seen) in [13].

Software Mockup Testing: In a 2 2 laboratoryexperiment, 48 participants with differing levelsof computer SE (24 low, 24 high) and of different

150 IEEE TRANSACTIONS ON PROFESSIONAL COMMUNICATION, VOL. 56, NO. 2, JUNE 2013

Fig. 8. Prototype B—Passenger selection (left) and options (right) screens.

ages (24 over and 24 under the age of 55)interacted with the three prototypes (within-design,balanced order). The interfaces were presentedto the participants on a touchscreen. For eachprototype, the participants were given four fictitioustravel stories and tasks based on the use casesdescribed before. The time needed to completethe tasks was recorded in the logfiles along withthe user interactions and errors. In addition, theparticipants were encouraged to verbalize theirthoughts and feelings via think-aloud techniquesto track their subjective experiences during thepurchase process. At the end of the experiment,the participants were required to complete a shortusability questionnaire on their system preferencesas well as a short semantic differential basedon the AttrakDiff questionnaire to measure thehedonic qualities of the three prototypes [14].After the experiment, we conducted interviewswith the participants to acquire insights intotheir explicit knowledge of the buying processes.During these interviews, the participants alsohad the opportunity to give reasons for theirstated preferences and to freely formulate theirexperiences with the prototypes.

Repeated measures analysis of variance indicatedsignificant effects both for between-subjectsfactors and for within-subjects factors.Firstly, the amount of time and numberof user interactions per trial were higherfor participants with low computer literacy

2,44 11.85 0.001 η 0.35 and olderparticipants 2,44 7.96 0.01 η 0.27with no significant interaction. Second, theparticipants needed less time and fewer screeninteractions to complete the tasks with prototype

C (train metaphor) than with the other prototypes4,42 3.79 0.05 η 0.27 .

Errors or difficulties in use were encountered almostexclusively in the selection of passengers—theparticipants chose the wrong kind or wrong numberof passengers or selected a wrong travel card (suchas forgetting or not finding the discount option,choosing the wrong discount option, or mixing upindividual and family travel cards) in a surprisinglyfrequent number of cases when using prototypes Aand B. Virtually no such errors were recorded whenthe participants used prototype C. (See Table I.)

Usability ratings (as well as user preference)for prototype C were independent of individualcomputer SE levels (see Fig. 10), unlike prototypesA and B, which were assessed less positivelyby participants with lower computer SE scores( 1 9.42, 0.01 η 0.17 . Prototype Creceived better usability ratings than prototypes Aand B 2 18.11 0.001 η 0.28 .

The critical problems with these prototypes werereflected in the comments made by participants(and recorded during the experiment). PrototypesA and B both triggered negative experiences withpassenger selection: “I can’t press the ‘Complete’button until I’ve chosen all the passengers.”, “Oh no,here we go again! If I select two adults, I can’t addthe child.” Other statements related to navigationproblems (“There’s so much information on thescreen, but not what I need to know.”) or difficultiesgrasping the general software concept (“Typicaldeveloper logic ”).

Older participants were particularly prone tostopping during the difficult cases and asking(or looking at) the test coordinator for support.

SIEBENHANDL et al.: UCD APPROACH TO SELF-SERVICE TVMS 151

Fig. 9. Prototype C—Passenger selection (left) and options (right) screens.

TABLE IERRORS RECORDED IN USE CASE 2 (FAMILY)

Prototypes A and B also elicited verbal reactionsamong participants in this age group that indicatedan aversion to using this kind of TVM in the future:“I’d rather go by car!”, “I’d rather stand in line at thecounter behind ten other people.”, “Very confusing.If this were real life, I would walk away from themachine now.”

A few of the younger, computer literate participantsindicated a preference for prototype B, even thoughthey had encountered difficulties with it in theinitial trials: “It’s actually very easy.” The sametarget group described prototype A as “possibly toocomputer-like” or “complicated”. Nevertheless, theydid appreciate the simple approach to purchasinga ticket for a single journey offered by this system:“It’s simple for one person.” Comments on themoving train and graphical selection of passengerswere almost exclusively positive: “It’s fun”, “Neat, Ilike the train going around”, “Cool!”, “The figures arevery concise”. Only one participant did not like thebasic idea behind this system: “That’s nonsense—Iwant to buy a ticket, not play with a toy.”

Software Refinements: The results of the softwaretest indicated that prototype C (“Train Metaphor”)was perceived to be easy to use by people with

a generally low level of computer SE, and thatthis target group could be encouraged to considerusing the TVM option when confronted with sucha system. By applying an everyday graphicalmetaphor (selecting passengers by moving theminto a train), the user was guided through thepurchase process without a need for specificcomputer domain knowledge. These findingsled to the consortium’s decision to develop thisinterface further and drop the other two prototypes.Nonetheless, some aspects still needed furtherrefinement or were incomplete, such as theautomatic display of a choice of low-cost options forthe selected parameters or a personalization systemthat allowed users to select their most frequentroutes. The look and feel of the prototype were alsorevised to better reflect the ÖBB’s corporate identityand branding. (See Fig. 11.)

Integrated Software and Hardware Prototype: Ahardware development workshop for representativesof the specific target groups, hardware developers,industrial designers, and representatives of theÖBB was subsequently held at the DanubeUniversity Krems. The aim of this workshop wasnot only to find innovative ways of guaranteeingaccessibility and facilitating use of the new TVM,but also to discuss the feasibility of the newlyidentified possibilities. An industrial designerhired by the ÖBB subsequently designed the TVMprototype in line with the requirements catalog andideas generated at this workshop. The machinewas then built by the hardware manufacturers inthe consortium and featured two special hardwareelements designed to increase accessibility andefficiency (see Fig. 12): a QR code reader used notonly by the personalization system but also toallow the user to quickly buy the same (or return

152 IEEE TRANSACTIONS ON PROFESSIONAL COMMUNICATION, VOL. 56, NO. 2, JUNE 2013

Fig. 10. Usability ratings (left) and semantic differential (right) for the three prototypes.

ticket to a) ticket by simply scanning the original;a very large vertically mounted touchscreen thatautomatically adjusts the active part of the displayto the height of the user (inspired by [15]). Theintegrated prototype and refined software were usedfor the final round of user tests

Results of User Testing: Similar to the mockuptesting described before, a 2 2 laboratoryexperiment (within-design, balanced order) wasconducted with 48 participants with differing levelsof computer SE (24 low, 24 high) and of differentages (24 over and 24 under the age of 55). Theparticipants were given the same four fictitioustravel stories used in the mockup tests, but thistime, they interacted with the TVM prototype.They also had to solve one additional task: theywere handed a ticket with a QR code and toldto buy the return ticket for this route. The samemeasurements used in the software mockup testingwere also taken in this experiment (time neededto solve the tasks and user interactions/errors).In addition to the short usability questionnaire,participants were also asked to complete anenhanced version of the semantic differential beforethe postexperiment interview. Since computer SE

is dependent on the subjective feeling of havingmade progress during training [4], the participantswere asked to rate their confidence of being ableto successfully buy a valid ticket before and afterthe trials. The SE questionnaire used was createdspecifically for this experiment and was based onthe approach described by Bandura [16]. Finally,in a very similar approach to the product reactioncard method [17] used to assess brands andproducts [18], the participants were given a list of50 different adjectives and were asked to select thefive words that they felt best described the TVM.

The observations of the test participants’ behaviorrevealed some critical issues with the verticallymounted touchscreen. The most common suchissues were a parallax error (in 25 cases, almosthalf the participants in the test), followed by aninadvertent touching of the screen with the heel ofthe hand (in 16 cases). In addition, some testersunintentionally adjusted the height of the screen,since many of them did not realize that heightadjustment was a deliberate function (“I got a realfright when the screen moved.”). As a consequence,participants had to adjust their posture to the(usually too low) position of the screen, causing

SIEBENHANDL et al.: UCD APPROACH TO SELF-SERVICE TVMS 153

Fig. 11. Refined software—passenger selection (left) and options (right) screens.

additional parallax and typing errors (18 cases).Virtually none of the testers recognized the reasonfor this problem, assuming instead that either theywere at fault or that there were still some basicproblems with this version of the software.

The problems encountered with thepersonalization/quick purchase functionresulted primarily from the fact that the testerswere not familiar with or unable to find the QR codereader. (See Fig. 13.) A few of them associated thecode with the mobile phone function: “I have totake a photo of that with an iPhone. But I haven’t gotone!” Half the testers looked for the QR code on thescreen first, while many of them either ignored theslot containing the reader or confused it with theticket dispenser and tried to use other hardwarecomponents to read the code: the slot for banknoteswas the most common choice (17 people), butsome participants also tried to hold the ticket tothe camera (7), the screen (5), or the debit/creditcard payment slot (4). Nine people were unable toposition the ticket correctly, usually holding it toohigh. Several users reported that they would havebeen able to find the QR code reader quicker if thesymbol displayed at the top of the reader had beena QR symbol instead of a bar-code symbol. Only 17testers were able to use the QR code without anyassistance.

Despite these difficulties, all of the participantssaid in the postexperiment interviews that theywere extremely impressed with the new hardwarecomponents. They praised the QR personalizationfunction for its potential time-saving capacity andwelcomed the reduced typing requirements. Atthe same time, some participants expressed datasecurity concerns, particularly since the machine

would allow the ÖBB to monitor and store mobilitybehavior data. It was also pointed out that featuresof this kind would have to be clearly communicatedin the future because—with the exception of “earlyadopters”—most users would not find them out forthemselves.

The only problems explicitly mentioned with regardto the touchscreen were potential hygiene issuesand the general lack of feedback provided on thescreen. In the interviews, the participants were verypositive about the height adjustment option—butonly after the function had been explained to themin detail: “That must have been developed by realexperts—but you wouldn’t have known unless youwere told.” A further perceived advantage of thesystem was the fact that people with poor eyesightcould get very close to the screen without having tobend over.

One of the major points learned about the hardwarefeatures from the user tests was that they couldeasily interfere negatively with normal usage ofthe TVM. The weaknesses in the touchscreenand the software used to control the heightadjustment (which could, at best, be described as“experimental,” lacked reliable face recognition andwas very susceptible to changing light conditions)caused severe problems and confusion by notworking properly in every trial. If this systemwas actually to be implemented in a real TVM, itwould need to be guaranteed that the hardwarewas 100% dependable. The problems encounteredby the participants in using the new hardwareclearly indicate that the usability (and especiallythe visibility) of such systems has to be enhanced.Providing additional feedback and information,such as through acoustical signals and/or onscreen

154 IEEE TRANSACTIONS ON PROFESSIONAL COMMUNICATION, VOL. 56, NO. 2, JUNE 2013

Fig. 12. TVM currently in use (left) and created by the project consortium (right).

Fig. 13. Correct usage of the QR code scanner.

explanations, might also facilitate interaction withthe more innovative hardware components.

In addition to the observed results, the statisticalanalysis revealed that the starting valuesfor the ticket purchase SE scale (measuredbefore solving the tasks) were significantlyhigher for participants with a high level ofcomputer SE (CSE) than for those with lowcomputer SE 1 9.17 0.01 η 0.28 .Yet, these values also increased significantlyafter the participants had used the TVM

1, 49 29.41 0.001 η 0.38 . Theonly exception here was the group of elderlypeople with low CSE (significant interaction:

Fig. 14. Changes in ticket purchase SE for the testgroup.

time of measurement*age*CSE, 1,49 4.70,0.05 η 0.09). (See Fig. 14.) Overall, the

machine did not (as yet) succeed in conveyinga sense of success among the members of thisparticularly relevant target group and raising theirconfidence in their own abilities.

The semantic differential also produced differentresults within the different groups. In general, theTVM was assessed most positively by older people

SIEBENHANDL et al.: UCD APPROACH TO SELF-SERVICE TVMS 155

Fig. 15. Semantic differential.

with a higher level of CSE and most negativelyby older people with a lower level of CSE. Thelatter group’s more negative impression of themachine is also reflected in its choice of descriptiveterms: cumbersome, complicated, stressful, notself-explanatory. (See Fig. 15.) These experiencesare reflected by the statements collected in thethink-aloud protocols: “I would get nervous nowif people were waiting behind me.” or “How did Imanage that? How did I get the two adults into the

train?” Touchscreen issues seemed to be partly toblame for the unsatisfactory experiences. One usereven gave up at the end of the trial after failing to hitthe “Buy Ticket” button as a result of the parallaxerror: “It just doesn’t work! Is there somethingelse I still have to do? I don’t know what to donext!”. Furthermore, the difficulties encountered inselecting the different offers—a point criticized bythe majority of the participants—seemed to add tothe impression that the system was complicated:

156 IEEE TRANSACTIONS ON PROFESSIONAL COMMUNICATION, VOL. 56, NO. 2, JUNE 2013

“Cut out the ÖBB jargon! Get rid of their internalcodes!”

Of the list of 50 characteristics provided to theparticipants with the request to select the five thatthey felt best described the TVM (product reactioncard method), the following were most frequentlyselected: structured process (17), well thoughtthrough (15), usable (14), satisfying (13), innovative(9), advanced (9), useful (8), comprehensible (7),accessible (7), helpful (6). It is also importantto note that the participants did not opt for themore emotional descriptors, but instead chose toemphasize the more practical and efficient aspectsof the TVM. As one person put it: “I can get myticket—and that’s what I want.”

CONCLUSIONS, LIMITATIONS, AND SUGGESTIONSFOR FUTURE RESEARCH

What can be learned from the INNOMAT case studyfor the development of other self-service systems?In this final section, we will begin by derivingconclusions from our case study and discussing itslimitations before going on to delineate suggestionsfor further research.

Conclusions The INNOMAT project lasted morethan three years. The concepts and the design ofthe machine were developed by an interdisciplinaryteam of ergonomics, barrier-freedom, design,hardware, and software engineers and specialists.In potential realization terms, the participation offuture operator of the machine (in this case, theÖBB) within the project team and their feedbackwere of particular importance. From a developmentperspective, the interdisciplinarity of the teampresented a great opportunity, and confrontingdesigners, hardware developers, and softwarespecialists with the diverse needs and experiencesof different user groups helped us to realize noveltechnical and conceptual solutions and overcomelimiting factors.

An important challenge for the design of publicinformation and self-service systems like a TVM isthe heterogeneity of possible users [19] and themanifold user demands. This is why the INNOMATproject chose to follow the UCD approach. In ouropinion, the UCD methods, which helped us to getto know the context of use, merit specific mentionhere: The participant observations, interviews,and video analyses allowed us to identify variousneeds, barriers, fears, and requests that neededto be taken into account. The design process wassupported by explicating the user needs and by

focusing on frequently neglected user groups.We found that in a system designed for differentuser groups, some of the many requirementswere clearly contradictory (for example, althoughusers with little technological literacy argued for astep-by-step system that guides users through thewhole process, expert users also had to be takeninto consideration). It is necessary—and worththe effort—to search for reasonable compromises.In our case, we excluded some groups from ourtarget groups (for example, blind users wereexcluded early in the project) and developed novelapproaches to meet contradicting requirements (forexample, a QR scanner as a fast-access mode forexperienced users).

We know from recent studies that self-servicetechnologies have a particularly high impact onpeople with low affinity to technology; they posea barrier to their participation in daily life andfurther enlarge the digital divide in society. Ourcase study was able to show that low computer SE,especially in older people, influences interactionwith a novel TVM. However, we were also able toshow that by aligning the design of such a machinewith the narrative scripts of users, fewer interactionproblems and errors occurred than when thedesign was based on machine logic. Our semanticdifferential and qualitative interview data likewiseindicate that the UX was higher for a narrativeinteraction design.

We can therefore conclude that a narrative designapproach is of particular value for self-servicesystems [13]. However, adopting this new approachinstead of developing another variant of existinginteraction schemes did require a great deal ofnegotiation within the project team. It also requireda great deal of collaboration on the parts of thehuman–computer interaction and design partners.Despite being involved in the process from the verystart, the actual design ideas contributed by the(potential) users were of limited use in resolvingthe heterogeneous requirements for an inclusiveself-service terminal. Nevertheless, they did provideus with a lot of information on their thoughts andneeds, which broadened our knowledge about UXand ultimately influenced the design itself.

In our UCD approach, the development was brokendown into, planned, and launched in four phases,each with its own individual substeps: context ofuse, requirements analysis, design, and evaluation.While all of the process steps were ultimatelycompleted as planned, some adaptations did provenecessary along the way, especially with regard to

SIEBENHANDL et al.: UCD APPROACH TO SELF-SERVICE TVMS 157

the user tests. The project team wanted to covera broad cross section of the actual target group,which meant that in the course of the project,more than 250 people were involved in the focusgroups, interviews, and experiments. In line withthe theoretical plan, the first set of tests formedthe basis for the decision on the subsequentdesign, while the second experiment producedrecommendations for the implementation of theTVM. The tests were well planned, coherent, andconclusive, but, in practice, also very time andresource intensive. The experiments produced verydetailed data, which subsequently meant that agreat deal of time and effort was required to analyzeit, determine the key points for the next iteration,communicate this information to the designers andthe hardware/software developers, and negotiatethe necessary design decisions. For some projects,it might be worth considering a more flexibleapproach with smaller iterations—which are lesstime consuming and involve fewer resources—asthey can make it easier to steer the project andhandle detailed design questions. It also provedto be difficult to plan the UCD process prior tothe start of the project, as is suggested in [5].For example, the target groups for the prototypehad to be adapted during the CoU phase: moredesigns were realized than had been plannedin advance and programming took longer thanhad been scheduled (to mention just a few of thealterations that proved necessary). While a roughwork schedule is a definite necessity, adjustmentsand detailed planning should follow at a later stageand might require additional resources.

Although this case study was a research project,the development of the INNOMAT prototype was stillrestricted by some design limitations, a situationwhich might also arise in the development of otherself-service systems.

(1) Technical realization proved to be a limitingfactor for the design. For example, theplacement of hardware elements inside themachine limited the arrangement of externalelements.

(2) To lay the foundations for the futureproduction of the new generation of TVMin Austria, the interface had to mirror theexisting fare structure. From earlier research,we knew that the complexity of the railway farestructure (more than 1000 different types oftickets) is one of the major problems in the useof the current TVM [8]. This issue has beenrepeatedly raised in surveys, focus groups,

and other types of evaluations and, therefore,also had a strong influence on this projectas well, although it was essentially a matterof corporate strategy for the ÖBB and notdirectly linked to the development of the TVM.Nonetheless, the main goal of this projectwas to reduce complexity in the purchaseprocess without limiting the customer’s ticketoptions. Essentially, this meant that theproject team had to think of ways of usingthe TVM to at least reduce any need on thepart of the customer to take the fare structureinto consideration. The idea was to design a“smart” system that could limit the possiblefare options to those offers which were actuallyavailable and which represented the best valueoptions for the selected parameters and thecurrent train station. Users would then be ableto make their decisions based on prices andproduct descriptions. Although a remarkableeffort was made during the design process toreduce the number of options, the project teamcould basically only acknowledge customerdissatisfaction with the fare situation, since ithad neither the mandate nor the authority toinfluence the existing fare structure. However,the project team did at least document thepotential hurdles and dissatisfaction andcommunicate the problems to management atthe ÖBB during project meetings.

(3) Also related to the complex fare structure,the names used for the individual ticketsand a (lack of) knowledge of the terms andrestrictions of use, areas of validity, andsimilar issues remained potential hurdles—aswas seen in the final laboratory experiment.

(4) One other important factor that has tobe considered is the fact that TVMs areusually embedded in other sales channels.Interoperability can therefore only beguaranteed if there is broad support from theactual company and adequate coordinationand communication with other projects orneighboring systems (such as the websiteand the TVM sales channels). This raises thequestion of how to deal with incongruentsystems (such as the interfaces used on theinternet and the TVM) and with projects thatare insufficiently linked.

From a UX perspective, we would like to emphasize,in particular, that it is not simply sufficient fora self-service system to be “usable.” A positiveemotional experience is required to encouragewidespread use by a broad public. Prior studies

158 IEEE TRANSACTIONS ON PROFESSIONAL COMMUNICATION, VOL. 56, NO. 2, JUNE 2013

show that negative feelings toward the system andlow technological SE hinder access to self-serviceTVM [2]–[4], [8]. We therefore controlled in thiscase study for user technological SE and strived todevelop a system which raised positive emotionalreactions. To assess these, we observed how usersinteracted with the TVM and documented theiremotional behavior. We also asked the participantsto fill out a semantic differential which showed theemotional valence of the different interfaces. These“soft factors” of use may correlate with usability tosome extent, but they add the certain something toa self-service system that will make it a success.

Limitations This paper presents the INNOMATproject from the view of the HCI experts. If thesame story were to be told from the perspectives ofthe other project partners, the emphasis, details,and results would undoubtedly be different. Byhaving collaborated on the design instead of merelyevaluating it from an external perspective, we mightalso be biased by an insider’s view of the designdecisions in which we were involved. To reduce anysuch influence, we placed special emphasis on theobjectivity of the test methods.

As already mentioned, the original plans hadincluded more test and development iterations thanwere finally conducted. The actual duration of thedevelopment phases was highly underestimatedprior to the project. In order to complete theproject within the allotted timeframe and financialresources, two iterations had to be cut. Regrettably,the TVM could not ultimately be evaluated ina real-life setting as originally planned: Thedevelopment efforts this would have entailed

would have far exceeded the available scope andresources. Likewise, the integration of real cashverification elements and a debit/credit cardpayment function would have created legal andfinancial hurdles. Consequently, only previouslydefined use cases were realized and no links couldbe developed for actual, valid tickets. Accordingly, afield test will be conducted when a fully functioningTVM has been built.

Suggestions for Future Research The UCDapproach proved to be valuable for the developmentof self-service systems like TVM, and we wouldlike to encourage further research on self-servicesystems using this approach.

Some interesting questions remained open afterour case study. One such question is how frequentuse of the new TVM would influence technologicalSE and UX? We assume that by using the TVMmore often and becoming more familiar with theinteraction, users will produce fewer errors and canbuild up a positive feeling toward the machine, afactor which again might increase technological SE.

A second open question would be whether (andto what extent) the problems with the hardware(parallax error, height adjustment, QR scanner)influenced technological SE and UX. The impactof such problems would seem to be important indecisions as to whether self-service technologiescan be installed despite the existence of “minorbugs” or whether such features should be excludedif no guarantee can be given that they will function100% correctly. Our qualitative data suggest thelatter.

REFERENCES[1] ÉGALITÉ. (2006). Equality issues in everyday life in a telematics supported field of transport. [Online]. Available:

https://forschung.boku.ac.at/fis/suchen.projekt_uebersicht?sprache_in=en&menue_id_in=300&id_in=5863[2] Passenger Focus. (2008 Oct.). Buying a ticket at the station. Research on ticket machine use. [Online].

Available: http://www.passengerfocus.org.uk/research/publications/buying-a-ticket-at-the-station[3] Ö. Subasi, M. Leitner, N. Hoeller, A. Geven, and M. Tscheligi, “Designing accessible experiences for older

users: User requirement analysis for a railway ticketing portal,” Univ. Access Inf. Soc., vol. 10, pp. 391–402,2001.

[4] D. Struve and H. Wandke, “Video modeling for training older adults to use new technologies,” ACM Trans.Access. Comput., vol. 2, no. 1, pp. 1–24, 2009.

[5] M. Maguire, “Methods to support human-centred design,” Int. J. Human-Comput. Studies, vol. 55, no.4, pp. 587–634, 2001.

[6] B. Buxton, Sketching User Experiences: Getting the Design Right and the Right Design. San Mateo, CA:Morgan Kaufmann, 2007.

[7] M. Hassenzahl and N. Tractinsky, “User experience—A research agenda,” Behav. Inf. Technol., vol. 25, no. 2,pp. 91–97, 2006.

[8] G. Schreder, K. Siebenhandl, E. Mayr, and M. Smuc, E. Loos, L. Haddon, and E. Mante-Meijer, Eds., “Theticket machine challenge: Social inclusion by barrier-free ticket vending machines,” in Generational Use of NewMedia. Farnham, UK: Ashgate, 2012, pp. 129–148.

[9] G. Schreder, K. Siebenhandl, and E. Mayr, A. Holzinger and K. Miesenberger, Eds., “E-inclusion in publictransport: The role of self-efficacy,” in HCI Usability e-Inclusion. New York: Springer, 2009, pp. 301–311.

[10] A. Bandura, Self-Efficacy: The Exercise of Control. New York: Freeman, 1997.

SIEBENHANDL et al.: UCD APPROACH TO SELF-SERVICE TVMS 159

[11] S. J. Czaja, N. Charness, A. D. Fisk, C. Hertzog, S. N. Nair, W. A. Rogers, and J. Sharit, “Factors predictingthe use of technology: Findings from the Center for Research and Education on Aging and TechnologyEnhancement (CREATE),” Psychol. Aging, vol. 21, no. 2, pp. 333–352, 2006.

[12] L. Verhoef, Why Designers Can’t Understand Their Users. Utrecht, the Netherlands: Human Efficiency, 2007.[13] G. Schreder, E. Mayr, K. Siebenhandl, M. Smuc, and M. Nagl, “Narrative interaction as means for intuitive

public information systems,” Int. J. Public Inf. Syst., vol. 7, no. 3, pp. 143–149, 2011.[14] M. Hassenzahl, M. Burmester, and F. Koller, “AttrakDiff: Ein Fragebogen zur Messung wahrgenommener

hedonischer und pragmatischer qualität,” in Mensch & Computer 2003: Interaktion in Bewegung, G. Szwillusand J. Ziegler, Eds. Stuttgart, Germany: B. G. Teubner, 2003, pp. 187–196.

[15] S. Hagen and F. Sandnes, “Toward accessible self-service kiosks through intelligent user interfaces,” PersonalUbiquitous Comput., vol. 14, no. 8, pp. 715–721, 2010.

[16] A. Bandura, F. Pajares and T. Urdan, Eds., “Guide for creating self-efficacy scales,” in Self-Efficacy Beliefs ofAdolescents. Greenwich, UK: Information Age Publishing, 2006, pp. 307–338.

[17] J. Benedek and T. Miner, “Measuring desirability: New methods for evaluating desirability in a usability labsetting,” in Proc. Usability Prof. Assoc., 2002, pp. 8–12.

[18] T. Tullis and B. Albert, Measuring the User Experience. San Mateo, CA: Morgan Kaufmann, 2008.[19] B. Sundgren, “What is a public information system?,” Int.J. Public Inf. Syst., vol. 1, no. 1, pp. 81–99, 2005.

Karin Siebenhandl graduated in landscape engineering andreceived the Ph.D. degree in Landscape Engineering at theUniversity of Natural Resources and Life Sciences, Vienna,Austria, in 1997 and 2004, respectively. She is Head of theCenter for Cognition, Information and Management, DanubeUniversity Krems, Krems, Austria. Her research interests includeparticipatory information design, usability, and gender studies.

Günther Schreder graduated in psychology at ViennaUniversity, Vienna, Austria, in 2006. Since 2011, he has beena Research Associate at the Center for Cognition, Informationand Management, Danube University Krems, Krems, Austria.His research interests and teaching activities include HCI,participatory information design and narrative design.

Michael Smuc graduated in psychology in 2003 at ViennaUniversity, Vienna, Austria. He is Head of Research at the Centerfor Cognition, Information and Management, Danube UniversityKrems, Krems, Austria. His research interests focus on usabilitymethods, evaluation of Infoviz-tools, graph comprehension, anddynamic network analysis.

Eva Mayr graduated in psychology in 2004 at Vienna University,Vienna, Austria, and received the Ph.D. degree in appliedcognitive and media psychology at the University of Tuebingen,Tuebingen, Germany, in 2009. She is a Research Assistant atthe Center for Cognition, Information and Management, DanubeUniversity Krems, Krems, Austria. Her research interests focuson how new media technologies support cognitive processingand informal learning.

Manuel Nagl graduated in communication science at theUniversity of Vienna, Vienna, Austria, in 2004 and inneuroscience from the Vienna University and the MedicalUniversity of Vienna in 2010. He is a research associate at theCenter for Cognition, Information and Management, DanubeUniversity Krems, Krems, Austria. His research interests includeusability, information design, corporate communications, andnarrative design.