An Agile Framework for Modeling Smart City Business Ecosystems

Anne Faber1, Adrian Hernandez-Mendez1, Sven-Volker Rehm2 and Florian Matthes1

1Technical University of Munich, Boltzmannstrasse 3, 85748 Garching, Germany2WHU - Otto Beisheim School of Management, Burgplatz 2, 56179 Vallendar, Germany

{anne.faber, adrian.hernandez, matthes}@tum.de, sven.rehm@whu.edu

Keywords: Agile; Adaptation; Business Ecosystem; Smart City; Visualization

Abstract: Modeling business ecosystems enables ecosystem stakeholders to take better-informed decisions. In this paperwe present an agile framework for modeling a smart city business ecosystem. We follow a design science re-search approach to conceptualize the agile approach to manage ecosystem models and present the architectureof our framework as design artefact. During the design process, we evaluated ecosystem models that need toadapt with the emerging structures of the business ecosystem. The platform aims at a collaborative model-ing process, which empowers end-users to manage the business ecosystem models and underlying data. Theevaluation of the platform was conducted with industry partners as part of the presented smart city initative,indicating it usefulness when fulfilling modeling related tasks.

1 INTRODUCTION

Digitalization has long reached cities and is chang-ing urban mobility. Digital technologies are progres-sively integrated to vehicles, traffic systems, and in-frastructure. Internet of Things (IoT) devices for in-stance include sensors that provide information aboutoccupied or available parking slots. The created tech-nological affordances are thereby changing the mobil-ity opportunities and demands of travelers (Mitchell,2010). The variety of services enabled by digital tech-nologies currently ranges from providing timely in-formation on the traffic situation, to buying tickets forpublic transportation online, to car or bike sharing ser-vices, etc. The services usually comprise consumer-facing mobile applications that rest on digital plat-forms integrating various underlying services.

This technology-driven opportunity space createsa growing market that challenges established mobil-ity providers such as automotive OEMs, their tier 1to 3 parts suppliers, but also public transportationproviders. One challenge originates from technologystartups deploying new IoT technologies such as sen-sor technology, augmented reality or artificial intelli-gence to urban mobility. Tech giants such as Googleand Apple are also entering the mobility marketsglobally, by developing self-driving cars and push-ing autonomous driving (Etherington and Kolodny,2016), (Taylor, 2016). As a result, new businessecosystems are currently emerging around mobility

markets that are geographically focused on specificmetropolitan areas. Besides commercial mobilityproviders, cities, public institutions and their govern-ments are addressing these challenges as part of theirurban development policies concerning smart cityconcepts. They increasingly become actors within theemerging mobility business ecosystems.

Understanding the evolution process of such mo-bility ecosystems is instrumental for developing pub-lic policies, for taking strategic decisions about busi-ness and technology partnerships, or for identifyinggaps in the service provision to consumers (Basoleet al., 2015a). Hence, the proactive management ofthe business ecosystem is gaining relevance for firmsas well as city authorities (Basole et al., 2015a). Par-ticularly, firms have to adapt their own competenciesto their specific (part of the) ecosystem to achievecomplementarity (Leonardi, 2011), (Rehm and Goel,2017).

Our research contributes to this issue by providinga community-based, agile approach to model and vi-sualize smart city mobility business ecosystems. Webase our insights on own software engineering de-sign work and a field test of the developed system.The case study is part of a smart city initiative pur-sued by a European city with a population of morethan 2.5m in its urban area and more than 5.5m in itsmetropolitan region. The mobility ecosystem is an-ticipated to embrace more than 3.000 firms in the au-tomotive, traffic and logistics sectors residing in the

urban area and more than 18.000 firms in these sec-tors in the metropolitan region.

1.1 Problem Statement

Visualizing data is a widely used approach to de-rive value from data by spotting anomalies and cor-relations or identifying patterns and trends (Vartaket al., 2017). This holds true for the context of busi-ness ecosystems, as visualizations of ecosystems haveproven to enable ecosystem stakeholders to take bet-ter informed decisions (Basole et al., 2016), (Huhta-maki and Rubens, 2016), (Evans and Basole, 2016).In the context of visual decision support, visual an-alytic systems (VAS) have been proposed and eval-uated to leverage related benefits (Park and Basole,2016), (Park et al., 2016). These systems allow ad-dressing needs and demands of diverse user groupsthrough different views and types of visualizations(layouts). VAS system architecture comprises ele-ments for interaction of users, for interpreting thevisual input, and for generating meaningful reports(Park et al., 2016).

One success factor of visualizing ecosystems isthe availability of ecosystem data used for these vi-sualizations (Park et al., 2016). Ecosystem data com-prises (a) technology-related data, such as availableservices, technological standards and platforms, mon-itoring data sources, (b) business-related data, suchas information about service providers, their strate-gies, partnerships and offered solutions, cooperativeinitiatives, as well as (c) market-related data, such asregional coverage of services, user types (commuter,tourist etc.), or use patterns of mobile service apps.Ecosystem experts and data scientists together evalu-ate and interpret the data to create tailored visualiza-tions.

Research addressing ecosystem models and visu-alizations has used sets of data collected from com-mercial databases on business and economic data ordrawn from social or business media (Basole et al.,2015a), (Basole et al., 2015b). The required vari-ety of data sources effects extensive data collectionefforts, also requiring editorial revision of collecteddata. Data evaluation is therefore often executed onlyfor a specific timeframe, resting on static data sets.

As the structure of the mobility ecosystem in fo-cus is just emerging, the VAS ecosystem model, com-prising data model and view model which are usedto generate and visualize structures, must be adapt-able to address changing data sets. Regarding the datamodel, e.g., new service providers must be linked tothe right types of services or positioned in the marketbut can also constitute new types of firms or exhibit

new types of relationships that subsequently need tobe created in the data model. Regarding the viewmodel, in general-purpose VAS, visualization are of-ten not adaptable without high effort. Thus, the viewmodel needs to have the capability to include newstructures from the data model.

In addition to these aspects concerning datasources and technical requirements, the data collec-tion and editing process as well as the visualizationprocess face further challenges. For editing data,team-oriented approaches provide a way to cope withthe complexity and heterogeneity of data sources andbusiness/technology contexts to cover. As this editingprocess generates the input to the visualization pro-cess, both processes need to be linked within the VASin order to provide high flexibility for interacting andinterpreting with help of the visualization user inter-face, to define relevant key indicators and to createtailored reports.

1.2 Contribution

Our approach addresses the aforementioned chal-lenges by suggesting an agile approach to collabora-tively manage and adapt business ecosystem modelsand visualizations. We have developed a VAS we re-fer to as Business Ecosystem Explorer (BEEx), in-tended as a framework for understanding emergingstructures of smart city business ecosystems. This ag-ile framework allows to collaboratively aggregate andmap data about the ecosystem, define analytic struc-tures and create multiple types of views, thus provid-ing a customizable instrument for different ecosystemstakeholders as users of the system. Data about theecosystem allows to visualize the past development ofthe considered ecosystem and enables stakeholders toanalyze present structures, e.g., which company posi-tioned itself as key player within the ecosystem (Ba-sole et al., 2015b).We provide insights from the prototypical use of theframework developed by us in our case study of alarge European smart city initiative. We focus on theiterative and emergent character of the process to col-lect, edit and visualize data, including the feedbackloop stimulating review of gathered data and struc-tures, and re-formatting of visualizations to customizethem for different stakeholder needs. This feedbackloop is the source of several requirements concern-ing adaptability and collaborative use of the VAS(Leonardi, 2011), (Majchrzak et al., 2002). In re-sponse to that, we conceptualize the ecosystem mod-eling process by focusing on stakeholder roles as wellas visualization and interactivity aspects. After de-scribing our methodological approach, we present re-

lated research and basic features of our tools data andvisualization models. Then we describe the iterativeapproach to collaboratively manage and adapt busi-ness ecosystem models, highlighting features that al-low for an agile data management. We lastly provideconsiderations on limits and future research with re-spect to two possible use cases for ecosystem VAS.

2 APPROACH

We have adopted a design science research approach(Hevner et al., 2004), (Hevner and Chatterjee, 2010).This perspective suits our research context of a casestudy that is part of a European smart city initiative.In the metropolitan region in focus, the structures ofthe mobility ecosystems are currently emerging as ef-fected by public authorities re-evaluating their policyand regional development activities, and businessestrying to co-shape the evolving mobility ecosystem.Our design artefact is a visual analytic system (VAS),the Business Ecosystem Explorer, which is particu-larly instrumental to provide tailored visualizationsto different stakeholders supporting them in theirbusiness- or policy-related tasks and decisions. Ac-cording to Hevners three cycle view of design science(Hevner et al., 2004), (Hevner, 2007), our researchcontributes in the following ways to each of the cy-cles.

Relevance Cycle: The aforementioned challengeslinked to smart city and urban mobility provide theboundary conditions to our case study. In the evolv-ing ecosystem, application of the Business EcosystemExplorer is expected to improve the visibility and un-derstanding of ecosystem structures. We provide datafrom its prototypical use including expert (user) feed-back, in order to assess its utility in practice as well asthe viability of our approach.

Design Cycle: We use an established KnowledgeManagement System application development plat-form to implement the VAS as design artefact. Thisplatform contains functionality for data management,collaboration, and decision support. The system par-ticularly provides features to address both data andview model adaptation. We draw on recent visualiza-tion research to identify and implement five apt vi-sualization types. As in its current stage, the systemis able to adapt data and view models to implementfurther visualizations, we seek evaluation in the rel-evance cycle by conducting a prototypical field testand including expert feedback which might stimulatefurther developments.

Rigor Cycle: We have studied and evaluated lit-erature and existing artefacts in the domain of visu-

alizing and particularly, business ecosystem analysis.We draw on the current state of the art in modellingapproaches for data visualization and modeling. Inaddition to the Business Ecosystem Explorer as de-sign artefact, we contribute to the knowledge space byproposing an agile and collaborative process to man-age and adapt business ecosystem models.

3 RELATED WORK

3.1 Business Ecosystem Modeling

Since the conceptualization of business ecosystemsby James Moore in the mid-1990s, who definedit as a collection of interacting companies (Moore,1996), the concept has been widely studied (Guittardet al., 2015). Ecosystems are interconnected througha complex, global network of relationships (Basoleet al., 2015b). In a business ecosystem, firms takeon roles such as suppliers, distributors, outsourcingfirms, makers of related products or services, technol-ogy providers, and a host of other organizations (Ian-siti and Levien, 2004), all affecting the characteristicsand boundaries of the ecosystem. As firms continu-ously enter and leave the ecosystem (Park and Basole,2016), they constantly evolve and exhibit a dynamicstructure (Peltoniemi and Vuori, 2004). Research onbusiness ecosystems has recently highlighted the roleof novel challenges for ecosystem formation, includ-ing technology contexts, e.g., the Internet of Things(IoT) (Iyer and Basole, 2016) or policy contexts, e.g.,smart city (Visnjic et al., 2016). This has focused re-searchers attention on ecosystem modeling (Uchihiraet al., 2016). Current approaches focus on frame-works to grasp the scope of ecosystem complexity(Iyer and Basole, 2016), (Visnjic et al., 2016), or onvisualization to understand emerging structures andpatterns (Leonardi, 2011), (Iyer and Basole, 2016).

Ensuing previous research, our business ecosys-tem model takes into account both the static networkof entities (firms, technologies), and the dynamic net-work characteristics, i.e., the relationships betweenentities, and activities, all changing over time. Enti-ties comprise small firms, large corporations, univer-sities, research centers, public sector organizations,(...) other parties [and human actors], which influ-ence the system (Peltoniemi and Vuori, 2004). Theyare linked through a variety of different relationshiptypes. All elements are to be integrated into the visualanalytic system (VAS). Which entities and relation-ship types need to modelled depends on the require-ments put forward by the VAS users, i.e., the (busi-ness) stakeholders. Their needs and demands that de-

Figure 1: Visualization types of the Business Ecosystem Explorer (BEEx): a) Force-Directed Layout (FDL), b) Matrix Layout(MXL), c) Modified Ego-Network Layout (MEL), d) Radial Network / Chord Diagram (RCD), and e) Tree Map Layout (TML)

fine which (visual) views are relevant, and which in-sights are vital, are fundamental for generating andadapting the model.

3.2 Business Ecosystem Visualization

To foster understanding of business ecosystems, Parket al. have presented a VAS (Park et al., 2016) that ad-dresses three salient design requirements of the par-ticular problem context of supply chain ecosystems.Their work includes extensive research in the areas ofmodeling, visualizing and analyzing across differenttypes of business ecosystems (Peltoniemi and Vuori,2004), (Iyer and Basole, 2016), (Visnjic et al., 2016),(Evans and Basole, 2016), (Park and Basole, 2016),(Basole et al., 2016). The described VAS offers mul-tiple views within an integrated interface, which en-able users to interactively explore the supply network.The system additionally provides data-driven analyticcapabilities. The authors suggest and test five visual-ization types (layouts) to visualize the dynamic net-worked structures of their problem context (Fig. 1).

These layouts comprise Force-directed Layout(FDL), Tree Map Layout (TML), Matrix Layout(MXL), Radial Network/ Chord Diagram (RCD),

and Modified Ego-Network Layout (MEL). Each ofthese layouts provides for interactive features, suchas clicking, dragging, hovering, and filtering. We usethese layouts suggested by previous research to ap-proach the design of the VAS in our problem con-text. In addition to these visualization types furtherones exist, which focus on particular aspects and per-spectives on ecosystems, such as cumulative networkvisualization (Evans and Basole, 2016), or bi-centricdiagrams that visualize the relative positioning of twofocal firms (Park and Basole, 2016), (Basole et al.,2016).

Current research on ecosystems at large uses data-driven approaches, i.e., sets of data are collected fromcommercial databases on business and economic data,or drawn from social or business media (Evans andBasole, 2016), (Park and Basole, 2016). This ap-proach implies that the VAS user, e.g., the strategyteam of a company, needs to understand requirementsfor relevant sources of data that inform the ecosys-tem model, as well as questions that guide visual-izations. Accordingly, both model and visualizationsneed to be adaptive to host diverse business perspec-tives and intentions. For the use of the resulting VASit is thus plausible to assume that business or public

Figure 2: Agile process to collaboratively manage and adapt the business ecosystem model

authority users might create their own VAS instancethat focuses on distinct data sources and structures.In the following, we therefore highlight the processof ecosystem modelling.

4 AGILE APPROACH TOECOSYSTEM MODELING

This section describes the process and roles necessaryto collaboratively manage and adapt business ecosys-tem models, and to visualize the models. We envisiona process that is adaptable to different types of teams,or communities. On the one hand, enterprise internalworking groups can constitute the stakeholders of anecosystem modelling initiative. On the other hand, apublic ecosystem model and VAS can be conceivedthat serves both the public and policy makers. Inde-pendent of these use scenarios, several roles need tobe represented that we present in the following.

4.1 Agile Modeling Process and Roles

Our approach comprises three phases that constitutean iterative procedure to initially build, use, and revisean ecosystem model (Fig. 2) (Roth et al., 2014).

First, the build phase comprises activities to mo-tivate creation and use of the VAS, collection ofdata about/from the ecosystem, and carry out themodelling. Basic requirements for engaging into an

ecosystem modelling initiative stem from the corestakeholders such as an enterprises top managementor strategy team. Together with domain experts, e.g.,business case owners, specific questions about theecosystem, its development or structures are formu-lated. These mirror the business strategy that under-lies the initiative at large. (For reasons of simplic-ity, we apply the use scenario of an enterprise, butroles will correspond to use in a public policy makerscenario, too). These requirements are taken up bya team (role) we named Ecosystem Editorial Team.This group is responsible for collecting data and mod-elling. (It is conceivable that for public use, a sep-arate, public or third party funded editorial office iscreated that overlooks and eventually investigates on,ecosystem data sources).

For the initial instantiation, company internal in-formation systems are used as data sources, provid-ing already collected information about competitors,business partners, etc. Additionally, each stakeholdergroup is motivated to implement their specific knowl-edge documented locally and to communicate thesources used to gather information. Within the it-erations of the build phase, these sources are usedto collect continuous information, both manually butalso automatically from news feeds, blocks, etc., bothorchestrated by the Ecosystem Editorial Team. Fi-nally, the stakeholder groups should be kept moti-vated to contribute with information whenever pos-sible. (For public use, available data sources, both

international, such as Crunchbase1 or Anglelist2, andnational, for Germany e.g. Grunderszene3 or Bayern-International4, are used for the initial build phase.In the next iterations, the data gathering extends tocrowdsourced data provided by the stakeholder alsousing the provided metrics, visualizations, and re-ports. Also, automatic news feeds evaluation are in-cluded to enrich the data base. It is within the respon-sibility of the mentioned editorial office to supervisethis process.)

In this phase, each stakeholder group owns spe-cific requirements towards both, understanding theecosystem as well as the functioning and use of theVAS. In our case study, for the Business EcosystemExplorer prototype, requirements from several stake-holders groups were collected; each group providedparticular demands with regard to relevant entities,and creation of views. For instance, legal departmentrepresentatives were rather interested in legal formsand business relationships of business partners, whilea strategy team focused on platforms and technolo-gies related to ecosystem members and cooperativeinitiatives, to inform the search for potential futurebusiness partners. The requirements are inititally col-lected in workshops lead by the Ecosystem EditorialTeam with each stakeholder group. In a later phaseof the VAS, the requirements are gathered within theVAS.

Second, the use phase covers presentation (execu-tion) of the created model within different layouts, in-teractions between users and the layouts to analyzethe ecosystem, and feedback to the Editorial Teamin order to fine tune or revise the model. In ourcase study, the business ecosystem model is presentedthrough stakeholder-specific ecosystem views that in-clude metrics, e.g., key values about centrality or con-nectedness of an entity, different visualizations, andreports. Reports play an important role in the com-munication process (Roth et al., 2014) and for ex-planation, as they contain the interpretation of dataand visuals by the domain experts, top managementor business stakeholders as users. This interpretationat a specific point of time serves as input to furtheranalysis and revision, and helps to follow the emer-gence of structures or patterns at a later stage.

Third, the revision phase comprises the reflectionon achieved results and validity of the model/providedvisuals as well as the adaptation of model and re-quirements. In this phase, additional input from ex-ternal domain experts can be sought depending on up-

1https://www.crunchbase.com2https://angel.co/3http://www.gruenderszene.de/4http://www.bayern-international.de/en/

coming tasks (Basole et al., 2016). A key role in thisprocess is assumed by the Ecosystem Editorial Team,whose modelling expert members particularly requiresome domain knowledge about modelling. The teamshould be capable of managing the various stake-holder groups, deliver stakeholder-specific visualiza-tions and safeguard the process cycle. The task ofecosystem experts holding domain knowledge aboutbusiness or regional factors etc. is to collect infor-mation from the ecosystem and prepare it in the rightformat as content of the ecosystem model.

In our case study, as for the requirements putforward by the modelling process, we have imple-mented the ecosystem model and the VAS into an in-tegrated, adaptive collaborative work system. Thissystems ecosystem model is capable of integratingall stakeholders requirements and visualizations, andthus grows with increasing demands and solutions,e.g., visuals that comprise selected entities and re-lationships to answer specific questions about theecosystem. This integrative aspect is particularly rel-evant, if one unique system is to be developed for useby a larger ecosystem initiative, as is the case in ourcase study context of a smart city initiative. Mul-tiple stakeholders including public and private orga-nizations might then become users of the ecosystemmodel.

4.2 Adoption and Agility Aspects

From our field test, we have seen that it is vital toensure involvement of different stakeholder groups,and to keep them motivated to discuss and explore themodel across process cycles. In this respect we haveexperienced that it is helpful to present an early ver-sion of the business ecosystem model and visualiza-tions to address specific demands (Roth et al., 2013).Further, we have noticed that the availability of variedvisualizations can stimulate cross-contextual think-ing, which might lead to formulation of new key val-ues in interpreting the ecosystem, e.g. transitive rela-tionships that express the indirect closeness betweenentities.

As a result, each of the phases should be im-plemented as interactive and collaborative processesto enable early adaptation and validation of formu-lated requirements (Roth et al., 2013). Addition-ally, different stakeholders should be able to adaptand evolve the business ecosystem model and visual-izations without having software development skills.Thereby, the collaborative process must be supportedcontinuously by an information systems, which al-lows the end-users (i.e., users without software de-velopment skills) to modify the business ecosystem

model and visualizations at run-time (i.e., without theneed to stop and recompile the system to integratenew functionalities). In this sense, we use the term ag-ile to characterize the way in which the process fromrequirements definition to visualization and feedbackshould be managed.

5 DESIGN ARTEFACT: AN AGILEFRAMEWORK FORMODELING ECOSYSTEMS

5.1 The Hybrid Wiki Approach toCollaborative Work, andArchitecture

The previously described requirements to the ecosys-tem modelling process caused us to design an ag-ile framework for modeling ecosystems as integrated,adaptive collaborative work system supporting theevolution of both the model and its instances at run-time by stakeholders and ecosystem experts (i.e.,users without programming knowledge or skills). De-veloping and maintaining such collaborative environ-ments can be considered a difficult task. Recent re-search has suggested the Hybrid Wiki approach to ad-dress this challenge (Matthes et al., 2011). The Hy-brid Wiki approach has been used in different usecases and domains such as Enterprise ArchitectureManagement (Matthes and Neubert, 2011), (Bucklet al., 2009) and Collaborative Product Development(Shumaiev et al., 2014), (Hauder et al., 2013).

This framework rests on the Hybrid Wiki ap-proach as presented in (Reschenhofer et al., 2016) thatserves as Knowledge Management System applica-tion development platform and contains features fordata management as well as collaboration and deci-sion support. To create the business ecosystem modelwe use the Hybrid Wiki metamodel.

The Hybrid Wiki metamodel contains the follow-ing model building blocks: Workspace, Entity, Entity-Type, Attribute, and AttributeDefinition. These con-cepts structure the model inside a Workspace and cap-ture its current snapshot in a data-driven process (i.e.,bottom-up process). An Entity contains a collectionof Attributes, and the Attributes are stored as a key-value pair. The attributes have a name and can storemultiple values of different types, for example, stringsor references to other Entities. The user can createan attribute at run-time to capture structured informa-tion about an Entity. An EntityType allows users to

refer to a collection of similar Entities, e.g., organiza-tions, persons, amongst others. The EntityType con-sists of multiple AttributeDefinitions, which in turncontain multiple validators such as multiplicity val-idator, string value validator, and link value validator.Additionally, an Attribute and its values can be asso-ciated with validators for maintaining integrity con-straints.

The EntityType and AttributeDefinition areloosely coupled with Entity and Attribute respec-tively through their name. These elements specifysoft-constraints on the Entities and Attributes. Theuse of soft-constraints implies that the users arenot restrained by strict integrity constraints whilecapturing information in Entities and their Attributes.Therefore, the system can store a value violatingintegrity constraints as defined in the current model.

5.2 Business Ecosystem Explorer Views

The agile framework currently consists of five views;a landing page, detail view with company informa-tion, a relation view, a visualization overview, andseveral visualizations (Fig. 3). For all views, a menubar at the top of the page provides links to the otherviews available.

The landing page displays a list of top-level enti-ties as defined in the ecosystem model. For our casestudy, we have for instance used organizations thatare part of the smart city ecosystem. The detail viewlists a short text for entity description, attributes suchas location, key personnel, legal form, etc., and rela-tions of the entity such as group affiliation, financialor contractual dependencies, etc. The relation view isan implementation of the Radial Network/Chord Dia-gram (RCD). For a selected range of entities, it high-lights different types of relationships. When hoveringover an entity on the arc of the circle this entity andall associated relations and entities are emphasizedby a bold type, whereas the remaining relations andentities are grayed out. The visualization overviewpresents various options to select other specific visu-alization types as separate, more detailed visualiza-tion views (Fig. 3).

5.3 Business Ecosystem Model

The agile framework relies on (a) ecosystem datamodel, and (b) ecosystem view model, each withrespective features for creation and adaption. Bothmodels are encoded using the Hybrid Wiki meta-model. The ecosystem data model contains two Enti-tyTypes within the Hybrid Wiki metamodel: organi-zations and relations. Thereby, organizations are au-

Figure 3: Architecture of the Business Ecosystem Explorer (BEEx): a) A list of all companies, b) Detailed information ofthe ecosystem entities, c) One view focusing on different types of relations between ecosystem entities, and d) Overview ofavailable visualizations (see also Fig. 1 )

Figure 4: List of Categories and Types in BEEx and conversion to Tree Map Layout (TML)

tomotive OEMs, their suppliers, public institutions,mobility related projects, etc. The AttributeTypesin the organization are company name, abbreviation,logo, URL, short description, headquarter, CEO, cat-egory, and legal form. Additionally, the relations de-scribe different types of interaction between organi-zations, as information coming from companies web-pages, newsfeed, social media, etc. Therefore, the at-tributes of the relation are the type of relation, e.g.,cooperation, (partially) ownership, funding, etc., in-volved companies, the date, and the source.

The ecosystem view model is encoded as one En-tityType called visualizations. Each visualization hastwo elements: the first element is the link between thedata model and the visualizations. The second ele-ment is the specification of the visualizations, whichare described using the visual encodings of the vi-sual grammar Vega5, as presented and described in(Heer and Bostock, 2010) and (Satyanarayan et al.,2016). The Vega visualization grammar introducesa declarative language to describe visualizations in aJSON format. The main building blocks, which en-able static and dynamic visualization features, are a)data, including data but also all data transformations;b) marks, covering the basic description of the visual-ized symbols, e.g., shape and size of a node; c) scales,containing visual variables, such as the color coding;d) signals, including the different interaction options,e.g., dragging and dropping of entities; and in someinstances e) legends.

The proposed approach provides the feature of

5https://vega.github.io/vega/

adapting the models at runtime. An example withinthe business ecosystem scenario is the categorizationof organizations. The initial grouping into Car Man-ufacturers, Map Provider, Mobility Platforms, etc. ispictured in Fig. 4 in the background screenshot. Thislist shows the representation within the collaborativework system. The system provides the feature toadapt the categories at runtime: adding a new cate-gory, changing or deleting existing categories. All vi-sualizations using the groups of companies, such asthe tree map layout (as pictured in Fig. 4), the chorddiagram layout or the modified ego-network layout,are adapted at runtime as well.

5.4 FIRST EVALUATION OF THEAGILE FRAMEWORK

For the evaluation process, the agile framework ishosted on a University server accessible on two In-ternet sites.Initially, we conducted nine interviews in semi-structured form with nine different companies withintwo months, following (Weiss, 1995). We aimed ata balance between receiving a quantifiable evaluationbut also enabling interviewees to vary the depth of an-swer depending on own capabilities and willingness(Glaser and Laudel, 2010). The focus of these inter-views was to receive feedback regarding the existingbusiness ecosystem model. Therefore, a sample of at-tributes was presented, as well as some types of rela-tions and a visualization of the combination of entitiesthrough relations in the form of a force-directed lay-

out. All interviewees stated that the business ecosys-tem model supported them in understanding the re-lations within the presented business ecosystem andthat their knowledge of this ecosystem was increased.Some suggested immediately additional scenarios, forexample, the patent management in the pharmaceuti-cal industry.We used the interviews’ results to update the exist-ing prototype. As a next step, we conducted threein-depth interviews with additional three companies.To obtain a wider range of opinions, we selected threecompanies of different fields of activity. Namely, anautomotive OEM, a publicly funded non-research in-stitution and a software company with main businessarea addresses the connected mobility ecosystem. Allinterview partners were actively modeling their busi-ness ecosystem and all stated their perceived limita-tions with the current in use tools (mainly Microsoftproducts in connection with CRM tools in use). Ad-ditionally, two companies confirmed that differentstakeholders within in the enterprise have differentviewpoints towards the enterprise’s business ecosys-tem. All companies agreed that the prototype fostersthe understanding of the presented ecosystem and twocontinued that it would be interesting to use such atool within in their enterprise to collaboratively man-age the business ecosystem evolution.

6 CRITICAL REFLEXION

As visualizations help stakeholders to better un-derstand data, the here presented visual analytic sys-tem links the data model and the view model. Thisenables a dynamic mapping of the view model andthe ever changing data model without high develop-ing efforts.

Nevertheless, within the VAS different stakehold-ers performing various roles are neccessary. Thismeans, on the one hand, the business owner anddomain experts have to provide their expert knowl-edge by providing their data sources and their re-quirements, but also technological experts within theEcosystem Editiorial Team providing visualisations,reports and metrics. The quality of the VAS dependshighly on the inclusing of these different stakehold-ers.

Additionally, as visualizations are data-driven, thebusiness ecosystem visualizations rely heavily on theavailability and quality of data. The data collec-tion must therefore fulfill quality requirements, whichneed to be more clearly defined in future research toassure data reliability and validity. To address thisdemand for structured data collection processes, ap-

proaches to data governance are needed. In caseof firm-internal usage, this is accomplished by theEcosystem Editorial Team. In public usage, the datagovernance role could be assumed by an Urban DataGovernance Board.

An extension to the visualization of ecosystems isinclusion of data about the actual use of services frommobile devices and digital platforms, and sensor datafrom the mobility infrastructure. This might open upnew options for identifying missing consumer-facingservices as for instance specific mobility services thatare not available in a particular region and identifica-tion of mobility providers and solutions that can closethis gap.

As a final limiting factor, it should be noticed thatthe presented agile framework prototype is not in itsfinal version as it iteratively evolves.

7 CONCLUSION

In this paper, we report from the field test of a pro-totypical visual analytic system (VAS) in context of asmart city mobility business ecosystem. We proposean agile approach to collaboratively manage and adaptbusiness ecosystem models and their visualizations.We provide insights from the use of our VAS designartefact, an agile framwork for modeling ecosystems,developed by us as part of a case study of a large Eu-ropean smart city initiative. The agile framwork isbased on an application development platform, whichprovides the feature of adapting the underlying dataand view models at runtime. Thereby, we address theneed to change the data and the view model accordingto the emerging structures of the ecosystem.

The results of our field test indicate that additionaliterations of the design cycle are required to extendthe functionality of the developed VAS tool. There-fore, we envision evaluating the proposed approachwithin further business ecosystem scenarios, such asthe medical and pharmaceutical ecosystem or the le-gal informatics ecosystem. As a tool improvementstep, we envision to adapt existing visualization andadd new ones based on feedback from industry part-ners. In addition, acceptance criteria for ecosystemVAS need to be defined, having in mind the two ma-jor use cases mentioned. Measuring the success ofecosystem visualizations in general is a further chal-lenge that will require a broader study of VAS adop-tion throughout various ecosystems.

ACKNOWLEDGEMENTS

This work is part of the TUM Living Lab ConnectedMobility (TUM LLCM) project and has been fundedby the Bavarian Ministry of Economic Affairs andMedia, Energy and Technology (StMWi) through theCenter Digitisation.Bavaria, an initiative of the Bavar-ian State Government.

REFERENCES

Basole, R. C., Huhtamaki, J., Still, K., and Russell, M. G.(2016). Visual decision support for business ecosys-tem analysis. Expert Systems with Applications,65:271–282.

Basole, R. C., Russell, M. G., Huhtamaki, J., and Rubens,N. (2015a). Understanding Mobile Ecosystem Dy-namics: A Data-Driven Approach. ACM Transactionson Management Information Systems, 6(2):1–32.

Basole, R. C., Russell, M. G., Huhtamaki, J., Rubens, N.,Still, K., and Park, H. (2015b). Understanding Busi-ness Ecosystem Dynamics: A Data-Driven Approach.ACM Transactions on Management Information Sys-tems, 6(2):1–32.

Buckl, S., Matthes, F., Neubert, C., and Schweda,C. M. (2009). A wiki-based approach to enter-prise architecture documentation and analysis. Ecis,19(2009):1476–1487.

Etherington, D. and Kolodny, L. (2016). Google’s self-driving car unit becomes Waymo. Technical report.

Evans, P. C. and Basole, R. C. (2016). Revealing the APIecosystem and enterprise strategy via visual analytics.Communications of the ACM, 59(2):26–28.

Glaser, J. and Laudel, G. (2010). Experteninterviews undqualitative Inhaltsanalyse : als Instrumente rekon-struierender Untersuchungen. VS Verlag fur Sozial-wiss, Wiesbaden.

Guittard, C., Schenk, E., and Burger-Helmchen, T. (2015).Crowdsourcing and the Evolution of a BusinessEcosystem. In Garrigos-Simon, F. J., , Gil-Pechuan, I.,, and Estelles-Miguel, S., editors, Advances in Crowd-sourcing, pages 49—-62. Springer International Pub-lishing.

Hauder, M., Roth, S., Matthes, F., Lau, A., and Matheis, H.(2013). Supporting Collaborative Product Develop-ment through Automated Interpretation of Artifacts.Proceedings of the BMSD 2013 Third InternationalSymposium on Business Modeling and Software De-sign, Noordwijkerhout, The Netherlands, 8-10 July2013, 19:151–156.

Heer, J. and Bostock, M. (2010). Declarative LanguageDesign for Interactive Visualization. IEEE Trans-actions on Visualization and Computer Graphics,16(6):1149–1156.

Hevner, A. and Chatterjee, S. (2010). Design Research inInformation Systems: Theory and Practice, volume 22of Integrated Series in Information Systems. SpringerUS, Boston, MA.

Hevner, A. R. (2007). A Three Cycle View of Design Sci-ence Research. Scandinavian Journal of InformationSystems, 19(2):87–92.

Hevner, A. R., March, S. T., Park, J., and Ram, S. (2004).Design Science in Information Systems Research.MIS Quarterly, 28(1):75–105.

Huhtamaki, J. and Rubens, N. (2016). Exploring innovationecosystems as networks: Four european cases. Pro-ceedings of the Annual Hawaii International Confer-ence on System Sciences, 2016-March:4505–4514.

Iansiti, M. and Levien, R. (2004). Strategy as Ecology. Har-vard Business Review, 82(3).

Iyer, B. R. and Basole, R. C. (2016). Visualization to un-derstand ecosystems. Communications of the ACM,59(11):27–30.

Leonardi, P. M. (2011). When Flexible Routines Meet Flex-ible Technologies: Affordance, Constraint, and theImbrication of Human and Material Agencies. MISQuarterly, 35(1):147 – 167.

Majchrzak, A., Rice, R. E., Malhotra, A., King, N., andBa, S. (2002). Technology Adaption: The Caseof a Computer-Supported Inter-Organizational VirtualTeam. Journal of Futures Markets, 22(8):697–727.

Matthes, F. and Neubert, C. (2011). Wiki4EAM Using Hy-brid Wikis for Enterprise Architecture Management.7th International Symposium on Wikis and Open Col-laboration, page 1.

Matthes, F., Neubert, C., and Steinhoff, A. (2011). HybridWikis: Empowering Users To Collaboratively Struc-ture Information. 6th International Conference onSoftware and Data Technologies ICSOFT.

Mitchell, W. (2010). Reinventing the automobile : personalurban mobility for the 21st century. MassachusettsInstitute of Technology, Cambridge, Mass.

Moore, J. F. (1996). The Death of Competition: Leader-ship and Strategy in the Age of Business Ecosystems.Leadership, page 297.

Park, H. and Basole, R. C. (2016). Bicentric diagrams: De-sign and applications of a graph-based relational setvisualization technique. Decision Support Systems,84:64–77.

Park, H., Bellamy, M. A., and Basole, R. C. (2016). Vi-sual analytics for supply network management: Sys-tem design and evaluation. Decision Support Systems,91:89–102.

Peltoniemi, M. and Vuori, E. (2004). Business ecosystem asthe new approach to complex adaptive business envi-ronments. Proceedings of eBusiness Research Forum,pages 267–281.

Rehm, S.-V. and Goel, L. (2017). Using information sys-tems to achieve complementarity in sme innovationnetworks. Inf. Manage., 54(4):438–451.

Reschenhofer, T., Bhat, M., Hernandez-Mendez, A., andMatthes, F. (2016). Lessons learned in aligning dataand model evolution in collaborative information sys-tems. In Proceedings of the 38th International Confer-ence on Software Engineering Companion - ICSE ’16,pages 132–141, New York, New York, USA. ACMPress.

Roth, S., Hauder, M., and Matthes, F. (2013). Collaborativeevolution of enterprise architecture models. CEURWorkshop Proceedings, 1079:1–12.

Roth, S., Zec, M., and Matthes, F. (2014). Enterprise ar-chitecture visualization tool survey 2014. Technicalreport, Technische Universitat Munchen.

Satyanarayan, A., Russell, R., Hoffswell, J., and Heer, J.(2016). Reactive Vega: A Streaming Dataflow Archi-tecture for Declarative Interactive Visualization. IEEETransactions on Visualization and Computer Graph-ics, 22(1):659–668.

Shumaiev, K., Rehm, S.-V., and Reschenhofer, T. (2014). ISDesign Principles for Empowering Domain Experts inInnovation : Findings From Three Case Studies. Icis,pages 1–11.

Taylor, M. (2016). Apple confirms it is working on self-driving cars. Technical report.

Uchihira, N., Ishimatsu, H., and Inoue, K. (2016). IoT ser-vice business ecosystem design in a global, compet-itive, and collaborative environment. In 2016 Port-land International Conference on Management of En-gineering and Technology (PICMET), pages 1195–1201. IEEE.

Vartak, M., Huang, S., Siddiqui, T., Madden, S., andParameswaran, A. (2017). Towards VisualizationRecommendation Systems. ACM SIGMOD Record,45(4):34–39.

Visnjic, I., Neely, A., Cennamo, C., and Visnjic, N. (2016).Governing the City. California Management Review,59(1):109–140.

Weiss, R. (1995). Learning from strangers : the art andmethod of qualitative interview studies. Free Press,New York.