Enterprise Modelling and Information Systems Architectures Vol. 9, No. 1, June 2014 22 Ulrich Frank
Ulrich Frank
Enterprise Modelling: The Next Steps
Enterprise modelling is at the core of Information Systems and has been a subject of intensive research for about two decades. While the current state of the art shows signs of modest maturity, research is still facing substantial challenges. On the one hand, they relate to shortcomings of our current knowledge. On the other hand, they are related to opportunities of enterprise modelling that have not been sufficiently addressed so far. This paper presents a personal view of future research on enterprise modelling. It includes requests for solving underestimated problems and proposes additional topics that promise to promote enterprise models as more versatile tools for collaborative problem solving. In addition to that, the paper presents requests for (re-) organising research on enterprise modelling in order to increase the impact of the field.
1 Introduction at least one conceptual model of an organisa- tional action system with at least one conceptual It has been a wide-spread conviction for long model of a corresponding information system. that the complexity of large information systems Usually, but not necessarily, the various mod- recommends the use of models. Information sys- els that constitute an enterprise model are cre- tems are aimed at representing domains through ated with domain-specific modelling languages data that is accessible by prospective users. Rep- (DSML). To emphasise that enterprise models are resenting a–factual or aspired–domain cannot be intended to provide a medium both for fostering accomplished by modelling it directly. Instead, analysis and design tasks and for communication it comprises a twofold abstraction: We perceive across traditional professional barriers, the term a domain primarily through language, which in “multi-perspective enterprise model” has been turn reflects an abstraction over “objective” fea- introduced (Frank 1994). A multi-perspective en- tures of a domain. At the same time, using an terprise model is an enterprise model that em- information system requires an interface that cor- phasises accounting for multiple perspectives. responds to the language spoken in the targeted A perspective represents a specific professional domain. Therefore, the construction of informa- background that corresponds to cognitive dis- tion systems recommends the design of concep- positions, technical languages, specific goals and tual models. They do not only promise to reduce capabilities of prospective users (Frank 2013b). complexity by abstracting from ever changing pe- culiarities of technical infrastructures; they also In recent years, the term “enterprise architec- allow for getting prospective users involved in ture” has gained remarkable attention (Buckl et the analysis and design process. Exploiting the al. 2010; Land et al. 2009; Lankhorst 2005). The potential of information systems will often re- differences between enterprise model and enter- quire reorganising existing patterns of action— prise architecture are mainly related to the inten- sometimes in a radical way. Therefore, analysis ded audience. Enterprise modelling is aimed at and design of information systems should usually various groups of stakeholders that are involved be done conjointly with analysing and designing in planning, implementing, using and maintain- the organisational action system. The conception ing information systems. Therefore, enterprise of an enterprise model was developed to address models are supposed to offer a variety of cor- this demand. An enterprise model integrates responding abstractions. These include models Enterprise Modelling and Information Systems Architectures Vol. 9, No. 1, June 2014 Enterprise Modelling: The Next Steps 23
that serve as a foundation of software develop- At first, it represents a renunciation of the ori- ment. Therefore, the development of respective ginal approach to enterprise modelling, which DSML is a particular characteristic of enterprise was aimed at developing information systems modelling. Different from that, enterprise archi- from scratch. With respect to the complexity of tecture mainly targets IT management. There- today’s IT infrastructures and the fact that most fore, it puts less emphasis on the specification organisations will not develop substantial parts of DSML. Nevertheless, there is no fundamental of their information system on their own any- difference between both approaches. Instead, the more, this additional focus is certainly reasonable. abstractions used in enterprise architectures can Secondly, it is not only aimed at supporting the be seen as an integral part of more comprehens- design of IT infrastructures that are in line with ive enterprise models. In Information Systems, the corporate action system, but also at providing enterprise modelling has been a pivotal field of an instrument for IT management. Figure1 illus- research that has evolved over a period of more trates the representation of an enterprise model than 20 years (CIMOSA: Open system architec- through a set of interrelated diagrams that cor- ture for CIM 1993; Ferstl and Sinz 1998; Group respond to the current state of the art. To ensure 2009; Scheer 1992; Zachman 1987). It has produced integration, the partial models that are repres- various modelling frameworks, DSML as well as ented by the diagrams should be created with corresponding tools. The field has achieved a modelling languages that were specified with the remarkable degree of maturity which is indicated same meta modelling language and that share by the fact that enterprise modelling is part of common concepts (Frank 2011). many IS curricula—even though to different ex- While the abstractions covered by today’s en- tent. Nevertheless, there is still need for further terprise modelling methods arguably represent research to exploit the potential of enterprise relevant perspectives on an enterprise, a compre- modelling. In this paper I will point at relevant hensive representation of all aspects that may be shortcomings of the current state of the art in relevant for analysing, designing and managing order identify core elements of a future research a company together with its information sys- agenda. tem requires accounting for more context. While such a demand may look like an exaggeration to 2 The Need for More Context some, it is actually the consequent continuation of current practice: All professional activities in At the beginning, approaches to enterprise mod- a company are characterised by the use of con- elling were mainly focussed on developing high- ceptual abstractions, i.e., by a specific technical level frameworks to provide a common structure terminology and corresponding language games. or architecture of an enterprise and its informa- Reconstructing these terminologies through ad- tion system (CIMOSA: Open system architecture ditional DSML would not only enable further use for CIM 1993; Scheer 1992; Zachman 1987). Apart scenarios, it would also enrich existing models from using general purpose modelling languages with additional context. Context does not only (GPML) like the ERM and the UML, the devel- refer to the topics that are represented in an en- opment of DSML was mainly aimed at business terprise model. It also refers to the context in process modelling. Later, DSML were created for which the development and use of models occur. modelling strategies, organisational structures On the one hand, this kind of context includes or generic resources. In recent years, some ap- specific methods for enterprise modelling. On proaches have evolved that are aimed at DSML the other hand, it refers to organisational and for modelling IT infrastructures and IT architec- managerial arrangements to foster an adequate tures. This focus is remarkable for two reasons. handling of enterprise models. Enterprise Modelling and Information Systems Architectures Vol. 9, No. 1, June 2014 24 Ulrich Frank
Goal System Diagram Sustainable sales Labor costs 1 3 Value Chain Diagram Increase Keep sustainabl labor e sales costs Inbound Outbound Marketing stable Services Logistics Operations Logistics Sales Customer Number of satisfaction sales agents 2 4 Sustainable Sustainable sales sales 1 1 Increase Increase customer number of satisfactio sales n Increase Increase agents sust. sales sust. in region sales in A region B Organizational Chart
Business Process Map Board of Directors
Order Management Technical Support Complaint Product A Handling Product Asset Market Private A Manageme Systems Research Customers nt
Product Corporate Corporate Order Management Complaint Controlling Planning Credits Customers Product B
Order Complaint Management Handling Product Product C C
< Agent >
amount not Inform Customer Order < Agent > < Agent > available declined
Order Check Delivery Delivery Confirm Order < Fleet Manager > recieved Capacity possible Business Process Diagram amount Check Logistics < Agent > available
Delivery at Decline Order IT Resource Diagram desired date not possible
Object Model
Scan Order Invoice SAP CO SAP FI SAP R&D PLM 3.7 Customer Engine 5.1 created: Date issued: Date P10 6.0 P10 6.0 P20 custID: String issues 0,* no: String s total() : Decimal
1,1 0,* 1,1 e d
sum() : Decimal s u e
l 1,1
c Document 0,1 0,1 acts as total() : Decimal d n n
acts as u n i DMS l i o
c i
0,1 s t
Archive n t i
1.9 l a
r 0,1
1,* u c ABAP x Person 1,* e i 2.0 s l v Organisation e
p OrderItem InvoiceItem r lastName: String r p e
S name: String A units: Integer units: Integer firstName: String 0,* 0,1 P MW: SAP MW: SAP sum() : Decimal sum() : Decimal A born: Date
S BL: SAP Netweaver DB: t ERC ECC o 0,*
a sex: Gender t
s MDM PLM ADDS d r
e age() : Integer
e AccountEntry t f a t e c a 0,* r
created: Date o l
Address d 1,1 r e
r amount: Decimal 0,* t
e OS
e Product OS street: String a t v c s r o u
UNIX HP UX city: String l name: String e l UNIX HP
S 1,1
C 0,*
country: String id: String P
P UX 1,1 H
A 1,1 inStock: Integer Account S x i salesPrice: Money n Dell Power id: String Sun SPARC U retailPrice: Money Edge T110 balance() : Decimal
Figure 1: Diagrams Representing Example Enterprise Model Enterprise Modelling and Information Systems Architectures Vol. 9, No. 1, June 2014 Enterprise Modelling: The Next Steps 25
2.1 Further Modelling Topics of an enterprise and are usually not subject of design processes. Nevertheless, markets are of The variety of topics that are handled in enter- crucial importance for successful action in an en- prises is enormous. Among those that have not terprise. Furthermore, markets are getting more been sufficiently addressed in enterprise mod- complex and contingent: Often, they expand on elling are products, production processes, pro- an international scale and may be very dynamic jects, markets and logistic. While product mod- in the sense that products are displaced by in- elling is an issue on its own, integrating product novations or that customer preferences change models with enterprise models makes sense for quickly. Therefore, integrating models of mar- various reasons. Products can be very complex kets with enterprise models promises to gain a and may demand for quick adaptations. At the more differentiated understanding of relevant same time, developing, producing and handling market forces and to develop a better founda- products relates to various aspects of an enter- tion for decision making. Similar to production prise that are usually part of an enterprise model: processes, logistic networks have been subject of goals, business processes, organisational units or optimisation efforts for long. The respective mod- software systems. More and more, products com- els, often designed to satisfy the requirements prise software or are constituted by software. In of Operations Research methods, are mainly fo- addition to that products are often bundled—with cussed on optimisation with respect to certain services and/or other products. Therefore, integ- goals. Integrating the respective modelling con- rating product models with enterprise models cepts with languages for enterprise modelling would enable additional analysis scenarios such would enable to enrich both enterprise models as checking the effect of changing a product on and logistic models with more relevant context. required skills and on business processes. There are numerous approaches to model production In addition to traditional topics, organisations are processes. They aim at developing algorithms confronted with new phenomena that may de- and approximation procedures for production mand for appropriate action. They include social planning, process scheduling and process control. networks, virtual enterprises, nomad employees Integrating respective models with enterprise and many more. Extending enterprise models models will often be not trivial, because they with models of these phenomena would foster are based on different modelling paradigms. At analysing and handling them. This would, how- the same time, including elaborate models of pro- ever, require new modelling concepts. Finally, duction processes in enterprise models promises enterprise models can be supplemented with con- various advantages, such as supporting the con- cepts that are related to important further aspects joint analysis of production processes and related of managerial decision making. These include accounting concepts, e.g., specialised cost and business processes or generating software for benefit concepts, concepts to design and analyse controlling production processes from respective indicator systems (Strecker et al. 2012) as well models. In an increasing number of organisa- as concepts for modelling organisational know- tions, projects play a key role. Integrating project ledge. Adding these concepts would make enter- models with enterprise models would support prise models and corresponding tools a versatile project management by providing meaningful instrument for management—both on the oper- links to organisational resources. Also, project ational and strategic level. At the same time, it modelling could benefit from existing approaches could serve as a valuable extension of enterprise to business process modelling and would allow software systems (see Sect. 4.1). to take advantage of similarities between pro- jects. Markets have not been part of enterprise Request: To make enterprise models a versatile models for an apparent reason: They are outside tool for supporting professional action in organ- Enterprise Modelling and Information Systems Architectures Vol. 9, No. 1, June 2014 26 Ulrich Frank
isations, research needs to widen the scope of on the instantiation of methods from metamod- modelling by adding further topics that also com- els, while further approaches are based on a prise concepts to support managerial decision combination of composition and instantiation. making. It seems that the field has reached a stage of mod- erate maturity, which is also indicated by the spe- 2.2 Method Construction cification of a respective ISO standard (ISO/IEC 2007). Modelling languages are an important founda- Nevertheless, there are some aspects that have tion of enterprise modelling, since they provide been widely neglected so far. At first, current a purposeful structuring of a domain. However, approaches to method engineering are mostly they are not sufficient for designing and using en- generic in the sense that they are not restricted terprise models. In addition to languages, there to particular domains, nor do they account for the is need for processes that guide the purposeful peculiarities of enterprise models. That leaves development, interpretation and use of respect- prospective developers and users of enterprise ive models. In other words: There is need for models with the demanding task of adapting gen- modelling methods. Due to the diversity of pro- eric concepts to the idiosyncrasies of particular jects that can benefit from conceptual models, organisations. Second, current approaches to it is evident that a given set of modelling meth- method engineering focus on the design of pro- ods cannot not fit all demands—except for the cess models and take the modelling language as price of oversimplification. This insight shifted given. However, the diversity of topics that can the focus on approaches that guide the develop- be reasonable subjects of enterprise models may ment of customised methods. The only chance also require to adapt or even create modelling to provide support for the conceptualisation of languages. While a number of tools support the a range of methods is to increase the level of specification of DSML and the realisation of cor- abstraction by searching for essential character- responding model editors, prospective users can istics shared by all modelling methods. Against expect only little guidance with designing a lan- this background, the emergence of method en- guage that fits its purpose. At the same time gineering as a new field of research is a reas- the design of DSML is especially demanding. Of- onable consequence in two respects: First, it ten, prospective users will not have an idea of is aimed at rich abstractions that cover a wide what they might expect from a DSML. As a con- range of modelling projects. Second, it makes sequence, requirements analysis is a remarkable use of the same paradigm that it suggests for challenge. In addition to that, design conflicts the field of conceptual modelling, too: The con- need to be handled. Also, the creation of the con- struction of particular methods should follow crete syntax requires a specific competence that an engineering approach, which—among other many language designers do not have. There- things—recommends accounting for linguistic fore, there is need for substantiated guidance to rigour, consistency and coherence as well as for reduce the risk of poorly designed modelling lan- the development of supportive tools. During guages. Currently, there are only few approaches the last 15 years, a plethora of method engineer- that offer respective support (Frank 2013a; Moody ing approaches—originating mostly in Require- 2009). Finally, method engineering is often based ments Engineering and Software Engineering— on two assumptions: First, a method is an arte- has evolved (Brinkkemper 1996; Ralyté et al. 2005, fact that is created through an engineering act. 2007). For an overview see Henderson-Sellers and Second, applying the method appropriately is Ralyté (2010). Some emphasise the construction pivotal for successful action. However, with re- of methods from reusable elements, others focus spect to successfully using a method, it is not Enterprise Modelling and Information Systems Architectures Vol. 9, No. 1, June 2014 Enterprise Modelling: The Next Steps 27
sufficient to restrict it to its explicit definition, as linguistic artefacts and additional organisa- i.e., to take a mere technical perspective. This tional/managerial measures that promote the ap- is for two interrelated reasons. First, a method propriate use of methods in practice. will usually not be based on a pure formal spe- cification. Instead, its conceptual and theoretical 3 The Need for More (Re) Use foundation as well as the process description re- The remarkable effort that is required to build quire interpretations that produce some degree elaborate enterprise models makes reuse of mod- of shared understanding. Second, for a method els and modelling concepts a pivotal issue for to work, it has to become an accepted orientation achieving higher productivity. At the same time, for individual and collaborative action. To sum- reuse can also contribute to model quality, if re- marise both aspects: A method needs to make usable artefacts are designed and evaluated with sense. From this point of view, a method can specific care. In addition to that reusable con- be regarded as a social construction that reflects cepts can serve to foster integration of those com- established patterns of professional action, ideas ponents that share them. Approaches to promote of professional values and aesthetics, organisa- reuse have been on the research agenda for long. tional culture, common beliefs as well as indi- The idea of reference enterprise models seems vidual interests. Against this background, we to be especially attractive. However, so far, re- can distinguish between a method as a linguistic use of enterprise modelling artefacts remained artefact, stressing a technical view, and a method on a modest level (Fettke and Loos 2007). There as an actual practice, stressing a more pragmatic are various reasons that contributed to this un- or organisational view. Therefore I intention- satisfactory situation. Two especially important ally avoid the term “method engineering” and reasons are related to modelling languages. On speak of “method construction” instead. This the one hand, current languages for enterprise is to express that a method is also constructed modelling lack concepts that enable reuse. On by those who use it, because it is shaped by ac- the other hand, the design of reusable DSML is tual interpretations and actions. A method as an facing a substantial challenge. artefact could be regarded as input or stimulus to trigger such a process. While for analytical 3.1 The Lack of Abstraction in Process reasons it may be useful to focus on methods Modelling mainly as linguistic artefacts, such a restricted view is certainly not sufficient with respect to a Taken the fact that business process modelling pragmatic objective such as improving efficiency has been a research subject for long, it seems sur- and quality of collaborative problem solving in prising that respective modelling languages are organisations. The benefit of methods for en- rather primitive in the sense that they do not al- terprise modelling will not only depend on the low for powerful abstractions. As a consequence, qualification of the involved stakeholders, but reuse of business process models remains on a also on certain aspects of the respective corpor- dramatically poor level. Since business process ate culture. It makes a clear difference, whether models play a pivotal role within enterprise mod- conceptual models are regarded as corporate as- els, this is a serious shortcoming. The follow- sets or as cost drivers with dubious outcome. ing scenario illustrates the problem. A company comprises a few tens of business process types Request: To promote the beneficial development including a core order management process type. and use of enterprise models it is required to A process type includes activity types. Various support the construction of respective modelling process types share similar activity types. Now methods that account for both, the conceptual two more specific order management process foundation of designing/customising methods types need to be implemented. For this purpose, Enterprise Modelling and Information Systems Architectures Vol. 9, No. 1, June 2014 28 Ulrich Frank
standard order management
no chance for contract
calculate prepare contract term&conditions contract prepared order received terms&conditions satisfactory
order management for under-specified order
no chance for contract
determine options calculate prepare contract request options terms&conditions term&conditions contract prepared received determined satisfactory
Figure 2: Example of extending a business process type
it would be most helpful to specialise the existing ways effect the original control flow (Frank 2012). order management process type. The example in Fig.2 illustrates this problem.
This would not only allow reusing the respective There are a few approaches in Software Engineer- model and corresponding software implementa- ing and process modelling which are aimed at a tions, it would also promote safe and convenient relaxed conception of specialisation of behaviour maintenance: Future changes of the core process (Schrefl and Stumptner 2002) or of “workflow type would be immediately effective in the spe- inheritance” (Aalst and Basten 2002). Other ap- cialised types, too. To satisfy the demand for proaches focus on analysing structural similarit- integrity, a respective concept of process special- ies of control flows to promote reuse through pro- isation would have to satisfy the substitutabil- cess variants (Koschmider and Oberweis 2007). ity constraint: Every instance of a process type However, the restrictions these approaches imply can act as an instance of the corresponding su- remain unsatisfactory (Frank 2012). At the same per process type without causing harm (Liskov time, the still growing relevance of efficiently and Wing 1994). The substitutability constraint creating and maintaining business process mod- is satisfied, if the extensions defined for special- els demands for abstractions that allow taking ised concepts are monotonic. This can be accom- advantage of similarities. plished fairly easy for static abstractions. How- Request: Future research should aim at concepts ever, for dynamic abstractions such as business of relaxed process specialisation—which may be process models adding further activity or event combined with instantiation—that promote reuse types cannot be monotonic, because it will al- without unacceptable restrictions. Enterprise Modelling and Information Systems Architectures Vol. 9, No. 1, June 2014 Enterprise Modelling: The Next Steps 29
While the lack of a sound concept of process spe- same time, DSML foster model integrity, because cialisation creates a serious problem, the current they prevent the creation of inconsistent mod- state of business process modelling is even more els to a certain extent. By featuring a domain- dreary. The above scenario would suggest to re- specific concrete syntax, they also promote model use an activity type that was defined already for comprehensibility. Against this background, it a certain business process type in a new process does not come as a surprise that DSML are re- type. However, this is not possible: Every busi- garded by many as the silver bullet of conceptual ness process type has to be designed from scratch modelling and model-driven software develop- using the basic concepts provided by today’s pro- ment. However, their construction is facing a cess modelling languages. Hence, an activity dilemma. The more a DSML is tuned to a spe- such as “prepare contract” cannot be specified cific domain, the better is its contribution to pro- as a reusable type. Instead it is yet another in- ductivity and integrity. However, the more spe- stance of a basic (meta) type like “activity” or cific a DSML is, the more unlikely it can be used “automated activity” that is distinguished from in a wide range of particular domains. Figure3 other primarily through its label. There are ap- illustrates the conflicting effects of semantics on proaches that focus on analysing labels in order range of reuse and productivity. to detect similar activity types (Dijkman et al. e n
2011). However, their contribution to reuse is i s a u G e
y limited: Instead of removing the mess, they try R t
i f v o i
t e coping with it. c l u a d c o S r
Request P : There is need for extending business l a i t process modelling languages and tools with the n e t o possibility to define and reuse activity and event P types. This request is not easy to satisfy. An activity type is not only defined by its internal structure and behaviour, but also by its context such as Level of (domain-specific) Semantics the event that triggers it or the events it pro- duces. Reuse will be possible only, if the context Figure 3: DSML: Illustration of Essential Design Conflict can be adapted to some extent. Therefore, the required concept of an activity type—and of an Some authors suggest to design DSML to the event type respectively—must abstract from the needs of particular organisations or even pro- context without compromising reusability too jects only (Kelly and Tolvanen 2008; Völter 2013). much. This recommendation is based on two assump- tions. First, the variety of organisations would 3.2 The Essential Conflict of Designing not allow for powerful DSML that fit all indi- DSML vidual requirements. Second, there is no need for further reuse, because creating and using a DSML are characterised by convincing advant- DSML in one particular project will usually pay ages (Kelly and Tolvanen 2008; Kleppe 2009; Völ- off already. Even though both assumptions may ter 2013). By providing domain-specific concepts, be valid to a certain extent, they are hardly con- they promote modelling (and programming) pro- vincing. There may be remarkable differences ductivity: Modellers are not forced anymore to between organisational actions systems and cor- reconstruct domain-level concepts from generic responding terminologies. However, it would be concepts such as “entity” or “attribute”. At the a sign of epistemological defeatism to deny the Enterprise Modelling and Information Systems Architectures Vol. 9, No. 1, June 2014 30 Ulrich Frank
chance of finding substantial commonalities. Fur- to create models that include concepts on the M0 thermore, it can be assumed that actual variety level. is also a result of in part arbitrary processes of organisational evolution, i.e., it is not a reflection Universal DSML Organisational Unit M2 of inevitable differences. Also, there is evidence Committee Position that technical languages work in a wide range Language Designer of organisations of a certain kind: The termin- ology used in textbooks will often fit an entire Specific DSML (Local „Dialect“) Quality Circle industry in the sense that it provides a respected Department M1 linguistic structure and serves as as common ref- Team Market Analyst Organisation Analyst erence for professionals. Nevertheless, there are organisation-specific adaptations of textbook ter- Particular Organisation Quality Circle minology. They include extensions, refinements Model Marketing Department Product Group PG 1 M0 and modifications, some of which may be ques- Market Research Market tionable. The argument that a DSML will already Manager Team Analyst MA2 pay off in single use scenarios is fine, but it could Figure 4: Illustration of multi-level modelling languages still be much more profitable, since a wider range of reuse would allow for much better econom- ies of scale. Therefore, it would be beneficial to Designing such language systems and corres- create hierarchies of DSML, where more specific ponding tools is far from trivial. It requires giv- ones are extensions and/or instantiations of more ing up prevalent architectures of modelling lan- general DSML. guages that feature a given set of classification layers (for respective approaches see Atkinson et Request : Research on DSML should aim at hier- al. (2009); Clark et al. (2008); Simonyi et al. (2006). archies of languages to enable both a wide range Instead, recursive language models such as the of reuse and customised languages for narrow “golden braid” architecture are more promising— domains. and more demanding at the same time, because Figure4 illustrates the idea of providing mod- they are not supported by most of today’s devel- elling languages on different classification lay- opment environments. Apart from that, design- ers. The highest level (“universal DSML”) corres- ing languages for enterprise modelling should ponds to textbook terminology. The concepts on account for a further issue. Current DSML are this level should be applicable to a wide range usually specified with metamodels. This is for of organisations, hence, promote economies of a good reason: On the one hand, this kind of scale. The universal DSML should be designed specification corresponds to a paradigm the mod- by experts that possess deep knowledge about elling community is familiar with. On the other the general domain as well as rich experience hand, it fosters the construction of corresponding with designing DSML. “Local” DSML represent tools, because a metamodel can be used as a con- more specific technical languages for organisa- ceptual foundation of a respective modelling tool. tion modelling that apply to a few organisations The semantics of DSML, e.g., the semantics of or to one only. They are designed by organisa- specialisation concepts, is usually based on the se- tion analysts that are familiar with the respective mantics of prevalent programming languages to domain. These local DSML that feature a graph- facilitate the transformation of models into code. ical notation much like the universal DSML can However, there are other language paradigms be used by authorised managers to specify par- and specification styles that might enrich enter- ticular organisational settings. The example also prise models. For instance, models designed with shows that there are cases where it makes sense logic-based languages allow for deduction could Enterprise Modelling and Information Systems Architectures Vol. 9, No. 1, June 2014 Enterprise Modelling: The Next Steps 31
enable more sophisticated approaches to analys- he could browse a graphical representation of the ing enterprise models. Since the semantics of corporate business process map, which shows all respective languages, which are typically found business process types, their interrelationships in Artificial Intelligence, is different from that and key performance indicators at a glance. He of DSML used for enterprise modelling today, could then select a business process type he is integrating them into enterprise modelling envir- interested in, study the model that describes its onments is a demanding task. execution and demand for further aggregate data that characterises it, such as the number of in- 4 The Need for Run-Time Use stances per month, average revenues etc. Also, he Originally, enterprise models like most other con- could select specific analysis views, e.g., a view ceptual models were intended for supporting the that associates a selected business process type creation of information systems only. However, with the IT resources it requires. If he was inter- it is obvious that they should be beneficial during ested in one particular business process type, he the entire life cycle of an information system. On could view the corresponding model. Then, he a more generic level, this issue is addressed by could leave the conceptual level and ask for the research on “models at runtime” (Blair et al. 2009). list of currently active business processes of this Multi-perspective enterprise models provide ab- type and inspect the state of the instances he is stractions of the enterprise that support decision interested in. In addition to that, advanced users making and other managerial tasks. Also, they could modify the enterprise system by changing can help people in organisations to develop a the enterprise model. The DSML, an enterprise deeper understanding of the action system, i.e., model is created with would help preventing ar- how their work is integrated into a bigger pic- bitrary modifications and hence contribute to sys- ture. In addition to that, enterprise models can tem integrity. An outline of a respective system, enable users to develop a better understanding referred to as “self-referential enterprise system” of the information system and its interplay with is presented in (Frank and Strecker 2009). Figure5 organisational patterns of action. illustrates the idea of integrating enterprise sys- tems with enterprise modelling environments. 4.1 Integrating Enterprise Models with Such a system would allow realising the vision of Enterprise Systems interactive models propagated by Krogstie (2007, In an ideal case, an enterprise modelling environ- p. 306): “The use of interactive models is about ment would be integrated and synchronised with discovering, externalising, capturing, expressing, a corresponding enterprise (software) system. representing, sharing and managing enterprise On the one hand, this would enrich enterprise knowledge.” In other words: It would be a con- systems not only with their conceptual founda- tribution to empowering people who work in tion, but also with a representation of the context and interact with organisations. The realisation they are supposed to operate in. On the other of self-referential enterprise systems does not hand, enterprise models would be supplemented only require developing further DSML, but also with corresponding instance populations. redesigning enterprise software systems. This would enable users to navigate from con- Request: To further exploit the potential of both cepts on various classification levels to instances— enterprise software systems and enterprise mod- et vice versa. The following scenario illustrates elling environments, research should aim at de- the benefit of drilling down from an enterprise veloping the foundations for integrating both model to instances. A department manager who kind of systems into a versatile tool for managing is new to a firm wants to get a better understand- and adapting an organisation and its information ing of the way business is done. For this purpose, system. Enterprise Modelling and Information Systems Architectures Vol. 9, No. 1, June 2014 32 Ulrich Frank
M2
M1
Is there a chance for outsourcing services?
M0
Is there a chance for outsourcing services?
Figure 5: Navigating an enterprise model and corresponding instances
4.2 Deficiencies of Prevalent classes in the enterprise system, from objects Programming Languages in the enterprise modelling environment. As a consequence, one would have to deal with the The integration of enterprise modelling envir- notorious problem of synchronising models and onments and enterprise systems does not only code. Figure6 illustrates how the M0 layer of require research on enterprise models and their representation. It also demands for system ar- modelling tools is overloaded and that concepts chitectures that cannot be satisfactorily accom- in modelling tools are located on a classification plished with prevalent programming languages. layer that is different from that of corresponding Integration implies common representations of concepts in an associated enterprise information shared concepts. In today’s modelling environ- system. ments, conceptual models are usually represen- ted by objects on the M0 level—even though they Recent developments in research on program- belong to the M1 or even a higher level. Over- ming language has produced (meta) program- loading the M0 level happens for a good reason: ming languages that were designed for creating Prevalent programming languages are restricted domain-specific programming languages. Lan- to the dichotomy of objects and classes. Hence, guages like XMF (Clark and Willans 2012; Clark there are no meta classes that were required to et al. 2008) are especially promising, since they specify classes—and that would allow treating allow for an arbitrary number of classification classes as objects, too. Therefore, a common levels, which enables a common representation representation of classes in both systems is not of models and respective code. Hence, modify- possible. Instead, the only way to associate a ing an enterprise model implies modifying the modelling environment with a corresponding en- respective part of the enterprise software simul- terprise system would be to generate code, i.e., taneously. Enterprise Modelling and Information Systems Architectures Vol. 9, No. 1, June 2014 Enterprise Modelling: The Next Steps 33
Enterprise Modeling Environment
MetaEntity Metamodel Editor name: String ... concepts of metamodelling language M1 M3 Position name: String averageSalary: Money concepts of metamodel M0 M2 availability: Level ...
Model Editor(s) Position name: String averageSalary: Money Enterprise Information System concepts of modelling language M1 M2 availability: Level ... Position name: String Programmer concepts of model M0 M1 Implementation Layers Programmer averageSalary: Money id: String availability: Level staffed: Boolean ...... M1 schema (classes, types) average: € 48.300 availability: #low pos3: Position p1: Programmer M0 runtime (objects, data) name = 'Programmer' staffed = true averageSalary = 48.300 ... corresponds to availability = #low ...
Figure 6: Mismatch of Classification Levels
Request: To advance the state of current model- characteristic for research to abstract from single ling environments, research needs to focus on cases and aim at constructions that work for an tools that overcome the limitations of current entire class of cases. Furthermore, applied re- programming languages. search is motivated by improving existing prac- tice with respect to certain goals. Unfortunately, 5 The Need for Collaboration the development of reference enterprise models and corresponding languages and tools requires Extending the scope of enterprise models and resources that are not available to a single re- developing them to an omnipresent represent- search institute. Furthermore, establishing and ation of organisations requires an amount of disseminating them in practice depends on eco- research and development that cannot be car- nomic and political aspects that are beyond the ried out by the current enterprise modelling com- abilities and intentions of academics. Against the munity alone. To advance the field, there is need background, it is obvious that there is need to for cross-disciplinary collaboration and for accu- bundle resources of research institutions. At the mulating resources. same time, it is necessary to get vendors of en- terprise software and prospective users involved. 5.1 The Importance of Bundling On the one hand, they need to be involved to sup- Resources port requirements analysis. On the other hand, The development of comprehensive enterprise using reference artefacts in practice is the only models will overburden most organisations. This way to promote their dissemination. Unfortu- is the case, too, for respective DSML. Reference nately, there are serious obstacles that impede enterprise models and wide-spread DSML are both bundling of research resources and involve- suited to effectively address this problem. Fur- ment of companies. Research is based on compet- thermore they would provide a foundation for ition and the idea of scientific progress. Collab- cross-organisational integration of action sys- oration of research institutes implies to give up tems and information systems, which could en- competition to a large extent. At the same time, able a tremendous boost of productivity. At the for reference artefacts to be beneficial they need same time, the development of reference artefacts to be consolidated—which may jeopardise sci- that combine a descriptive and a prescriptive ap- entific progress. While there are probably many proach constitute an attractive research goal: It is vendors and client organisations that would be Enterprise Modelling and Information Systems Architectures Vol. 9, No. 1, June 2014 34 Ulrich Frank
happy to use reference models, most of them will Enterprise modelling is not an end itself. Instead, likely be reluctant to participate in respective it is supposed to have a positive impact on a com- development projects, since the return on such pany’s economics and competitiveness. However, an investment would be hard to determine. Nev- assessing the costs of creating and maintaining ertheless, to promote the benefit of enterprise enterprise models, which may include the devel- models, reference artefacts that enable attractive opment of languages and tools, is not a trivial economies of scale are of pivotal relevance. task—and it is similarly challenging to determ- Request: There is need for initiatives to collabor- ine the benefits that can be contributed to the atively develop and disseminate reference arte- deployment of enterprise models. Apart from facts. They need to provide convincing incentives economic effects, the extensive use of enterprise both for academics and practitioners. models within an organisation may have an im- pact on how people perceive not only their tasks One of the prime examples of community-driven but also the entire organisation. The increase collaboration is free and open source software in transparency may have an effect on estab- (FOSS). Respective initiatives have successfully lished patterns of organisational power and may promoted collaboration of developers and users. require new approaches to managing organisa- Also, they led to software systems of surprising tions. Against this background it is obvious that quality, and, in some cases, to an impressive dis- enterprise modelling involves a wide range of semination. Inspired by the apparent success of demanding research questions that concern vari- some FOSS projects, corresponding “Open Mod- ous disciplines. Business and Administration in els” initiatives have been proposed (Frank and general is aimed at developing and improving ter- Strecker 2007) and inspired the creation of open minologies and methods that are suited to struc- model repositories (France et al. 2007), (www. ture and guide purposeful action in enterprises. openmodels.org, www.openmodels.at). While Various subfields, such as Financial Management, these repositories have triggered remarkable at- Accounting, Industrial Management, Logistics tention, there is still need for more active parti- could contribute to extend and deepen enterprise cipation. models. In Psychology, the interaction between cognitive models and external representations 5.2 Enterprise Models as Object and is a core research topic. Applied to enterprise Promoter of Cross-Disciplinary modelling this would include the question how Collaboration conceptual models effect individual and collect- Enterprise models are aimed at providing a me- ive decision making. That includes analysing the dium to foster communication between stake- impact of graphical notations on people’s ability holders with different professional background. to understand complex matters—and the develop- On the one hand that requires reconstructing ment of guidelines for designing notations that technical languages and professional patterns of fit certain cognitive styles. Both, from a psycho- problem solving. On the other hand it recom- logical and a sociological point of view, it would mends analysing how prospective users react be interesting to analyse how enterprise models upon the models they are presented with. That effect the social construction of reality, i.e., to includes concepts as well as their designation what extent people perceive the model as the en- and (graphical) representation. For using enter- terprise and what that means for the way they prise models effectively, software tools are man- (inter) act. Assuming that enterprise models may datory. Developing and integrating them with have a substantial effect on an organisation’s per- other enterprise software systems creates sub- formance implies challenging research questions stantial challenges for Software Engineering or— for economic studies that are not restricted to in other words—interesting research questions. enterprise models, but comprise the economics Enterprise Modelling and Information Systems Architectures Vol. 9, No. 1, June 2014 Enterprise Modelling: The Next Steps 35
of models and methods in general. Combining re- with more specific ones. Last but not least, enter- search results from various disciplines would not prise models provide a laboratory for learning, only contribute to advance our knowledge about because they convey a solid conceptual found- enterprise models and our ability to use them ef- ation of information systems and surrounding fectively, it is also suited to enrich the state of the action systems—and enable students to navigate art in the participating disciplines, since it would through an enterprise on different levels of de- integrate it with contributions from other fields. tail and abstraction. With respect to such a wide Therefore the following request could have a bet- and deep conception of enterprise modelling it is ter chance to succeed than yet another call for important not only to identify relevant steps of inter-disciplinary research. future research, but also to spread the word and Request: Advancing the field of enterprise model- encourage others to participate in joint projects. ling recommends to establish inter-disciplinary Further developing the field also requires to put research collaboration. more emphasis on assessing model artefacts. On the one hand that comprises the development of 6 Conclusion pragmatic criteria to evaluate models and model- ling languages with respect to an intended prac- In the past, enterprise modelling, though argu- tical use. On the other hand, it relates to assess- ably pointing at a core topic of Information Sys- ing the epistemological quality of model artefacts tems, has been subject of a rather small, special- as research results. Developing and applying re- ised research community. In Business and Ad- spective criteria is an important prerequisite of ministration it is regarded as too much focussed scientific competition and progress. on technical aspects by some, while some tradi- tionalist colleagues in Computer Science suspect In this paper I gave a personal account of the it of lacking formal rigour. However, enterprise topics we should address in the next years to modelling is more than analysing and designing advance our field. It is needless to say that other information systems—and it is certainly much relevant topics exist, too. I would hope that the more than drawing “‘bubbles and arrows”. Enter- requests presented in this paper contribute to a prise modelling is about conceptualising an im- discourse on our future research agenda. portant part of the world—as it actually is and as it might be. Hence, it requires knowledge about References how people (inter) act in organisations, how in- formation systems infrastructures are built—and der Aalst W., Basten T. (2002) Inheritance of the creativity to develop substantial images of Workflows – An Approach to Tackling Prob- attractive future worlds that comprise the pur- lems Related to Change. In: Theoretical Com- poseful construction and use of information sys- puter Science 270, p. 2002 tems. It is about how we perceive the world Atkinson C., Gutheil M., Kennel B. (2009) A Flex- we live and work in and how we think about it ible Infrastructure for Multilevel Language and might change it—alone and together with Engineering. In: IEEE Transactions on Soft- others. In addition to supporting collaboration ware Engineering 35(6), pp. 742–755 between stakeholders with different professional Blair G., Bencomo N., France R. B. (2009) backgrounds in organisations, enterprise models Models@ run.time: Computer. In: Computer may also serve as a medium and object of inter- 42(10), pp. 22–27 disciplinary research. At the same time, they Brinkkemper S. (1996) Method Engineering: En- are suited to foster the exchange between prac- gineering of Information Systems Develop- tice and academia, because they allow to integ- ment Methods and Tools. In: Information and rate more abstract representations of enterprises Software Technology 38(4), pp. 275–280 Enterprise Modelling and Information Systems Architectures Vol. 9, No. 1, June 2014 36 Ulrich Frank
Buckl S., Matthes F., Roth S., Schulz C., Schweda 2nd Edition. ICB-Research Report 43 C. (2010) A Conceptual Framework for Enter- Frank U. (2012) Specialisation in Business Pro- prise Architecture Design. In: Proper E., Lank- cess Modelling: Motivation, Approaches and horst M. M., Schönherr M., Barjis J., Over- Limitations. 51. University of Duisburg-Essen. beek S. (eds.) Trends in Enterprise Architec- Essen ture Research. Lecture Notes in Business In- Frank U. (2013a) Domain-Specific Modeling Lan- formation Processing Vol. 70. Springer, Ber- guages - Requirements Analysis and Design lin, Heidelberg and New York, pp. 44–56 Guidelines. In: Iris Reinhartz-Berger, Aron CIMOSA: Open system architecture for CIM. Sturm, Tony Clark, Yair Wand, Sholom Co- Springer, Berlin, Heidelberg and New York hen, Jorn Bettin (eds.) Domain Engineering: Clark T., Willans J. (2012) Software Language Product Lines, Conceptual Models, and Lan- Engineering with XMF and XModeler. In: guages. Springer, pp. 133–157 Mernik M. (ed.) Formal and Practical Aspects Frank U. (2013b) Multi-Perspective Enterprise of Domain-Specific Languages. Information Modeling: Foundational Concepts, Prospects Science Reference, pp. 311–340 and Future Research Challenges. In: Software Clark T., Sammut P., Willans J. (2008) Superlan- and Systems Modeling (in print) guages: Developing Languages and Applica- Frank U., Strecker S. (2007) Open Reference Mod- tions with XMF. Ceteva els – Community-driven Collaboration to Dijkman R., Dumas M., van Dongen B., Käärik Promote Development and Dissemination of R., Mendling J. (2011) Similarity of business Reference Models. In: Enterprise Modelling process models: Metrics and evaluation. In: and Information Systems Architectures 2(2), Inf. Syst. 36(2), pp. 498–516 pp. 32–41 Ferstl O. K., Sinz E. J. (1998) Modeling of Frank U., Strecker S. (2009) Beyond ERP Systems: business systems using the Semantic Object An Outline of Self-Referential Enterprise Sys- Model (SOM): A methodological framework. tems: Requirements, Conceptual Foundation In: Bernus P., Mertins K., Schmidt G. (eds.) and Design Options Handbook on Architectures of Information Group T. O. (2009) TOGAF Version 9. Van Haren, Systems. International Handbooks on Inform- Zaltbommel ation Systems Vol. 1. Springer, Berlin, Heidel- Henderson-Sellers B., Ralyté J. (2010) Situational berg and New York, pp. 339–358 Method Engineering: State-of-the-Art Re- Fettke P., Loos P. (eds.) Reference Modeling for view. In: Journal of Universal Computer Sci- Business Systems Analysis. Idea Group, Her- ence 16(3), pp. 424–478 shey Kelly S., Tolvanen J.-P. (2008) Domain-specific France R., Bieman J., Cheng B. (2007) Repository modeling: Enabling full code generation. for Model Driven Development (ReMoDD). Wiley-Interscience and IEEE Computer Soci- In: Kühne T. (ed.) Models in Software En- ety, Hoboken and N.J gineering. Lecture Notes in Computer Sci- Kleppe A. G. (2009) Software language engin- ence Vol. 4364. Springer Berlin Heidelberg, eering: Creating domain-specific languages pp. 311–317 using metamodels. Addison-Wesley, Upper Frank U. (1994) Multiperspektivische Un- Saddle River and NJ ternehmensmodellierung: Theoretischer Hin- Koschmider A., Oberweis A. (2007) How to de- tergrund und Entwurf einer objektorientier- tect semantic business process model vari- ten Entwicklungsumgebung. Oldenbourg, ants? In: Proceedings of the 2007 ACM sym- München posium on Applied computing. SAC ’07. Frank U. (2011) The MEMO Meta Modelling ACM, New York, NY and USA, pp. 1263–1264 Language (MML) and Language Architecture. Krogstie J. (2007) Modelling of the People, by the Enterprise Modelling and Information Systems Architectures Vol. 9, No. 1, June 2014 Enterprise Modelling: The Next Steps 37
People, for the People. In: Krogstie J., Opdahl ACM, pp. 451–464 A., Brinkkemper S. (eds.) Conceptual Mod- Strecker S., Frank U., Heise D., Kattenstroth H. elling in Information Systems Engineering. (2012) MetricM: a modeling method in sup- Springer Berlin Heidelberg, pp. 305–318 port of the reflective design and use of per- Land M. O., Waage M., Proper E., Cloo J., formance measurement systems. In: Inform- Steghuis C. (2009) Enterprise architecture: ation Systems and e-Business Management Creating value by informed governance. 10(2), pp. 241–276 Springer, Berlin Völter M. (2013) DSL Engineering: Designing, Lankhorst M. M. (2005) Enterprise architecture Implementing and Using Domain-Specific at work: Modelling, communication, and ana- Languages. dslbooks.org lysis. Springer, Berlin, Heidelberg and New Zachman J. A. (1987) A framework for inform- York ation systems architecture. In: IBM Systems Liskov B. H., Wing J. M. (1994) A Behavioral Journal 26(3), pp. 276–292 Notion of Subtyping. In: ACM Transactions on Programming Languages and Systems 16, Ulrich Frank pp. 1811–1841 Moody D. L. (2009) The “Physics” of Notations: Chair of Information Systems and Enterprise Toward a Scientific Basis for Constructing Modelling Visual Notations in Software Engineering. In: Institute for Computer Science and Business IEEE Transactions on Software Engineering Information Systems 35(6), pp. 756–779 University of Duisburg–Essen Ralyté J., Agerfalk P. J., Kraiem N. (2005) Universitätsstraße 9 Methods, Techniques and Tools to Support 45141 Essen Situation-Specific Requirements Engineering Germany Processes: SREP’05. University of Limerick, [email protected] Limerick Ralyté J., Brinkkemper S., Henderson-Sellers B. (2007) Situational method engineering: Fun- damentals and experiences. Proceedings of the IFIP WG 8.1 Working Conference, 12-14 September 2007, Geneva, Switzerland. IFIP - the International Federation for Information Processing Vol. 244. Springer, New York Scheer A.-W. (1992) Architecture of Integrated Information Systems: Foundations of Enter- prise Modelling. Springer, Berlin and New York Schrefl M., Stumptner M. (2002) Behavior- consistent specialization of object life cycles. In: ACM Transactions on Software Engineer- ing Methodologies 11(1), pp. 92–148 Simonyi C., Christerson M., Clifford S. (2006) Intentional software. In: Proceedings of the 21th Annual ACM SIGPLAN Conference on Object-Oriented Programming, Systems, Lan- guages, and Applications (OOPSLA 2006).