Hevner design-science


Published on

  • Be the first to comment

No Downloads
Total views
On SlideShare
From Embeds
Number of Embeds
Embeds 0
No embeds

No notes for slide

Hevner design-science

  1. 1. Hevner et al./Design Science in IS Research RESEARCH ESSAYDESIGN SCIENCE IN INFORMATIONSYSTEMS RESEARCH1By: Alan R. Hevner Sudha Ram Information Systems and Decision Management Information Systems Sciences Eller College of Business and Public College of Business Administration Administration University of Arizona University of South Florida Tucson, AZ 85721 Tampa, FL 33620 U.S.A. U.S.A. ram@bpa.arizona.edu ahevner@coba.usf.edu Salvatore T. March Own Graduate School of Management Abstract Vanderbilt University Nashville, TN 37203 Two paradigms characterize much of the research U.S.A. in the Information Systems discipline: behavioral Sal.March@owen.vanderbilt.edu science and design science. The behavioral- science paradigm seeks to develop and verify Jinsoo Park theories that explain or predict human or organi- College of Business Administration zational behavior. The design-science paradigm Korea University seeks to extend the boundaries of human and Seoul, 136-701 organizational capabilities by creating new and KOREA innovative artifacts. Both paradigms are founda- jinsoo.park@acm.org tional to the IS discipline, positioned as it is at the confluence of people, organizations, and techno- logy. Our objective is to describe the performance of design-science research in Information Sys- tems via a concise conceptual framework and clear guidelines for understanding, executing, and evaluating the research. In the design-science paradigm, knowledge and understanding of a problem domain and its solution are achieved in the building and application of the designed arti-1 fact. Three recent exemplars in the research Allen S. Lee was the accepting senior editor for thispaper. literature are used to demonstrate the application MIS Quarterly Vol. 28 No. 1, pp. 75-105/March 2004 75
  2. 2. Hevner et al./Design Science in IS Researchof these guidelines. We conclude with an analysis ness and efficiency of an organization. Theseof the challenges of performing high-quality theories impact and are impacted by designdesign-science research in the context of the decisions made with respect to the systembroader IS community. development methodology used and the functional capabilities, information contents, and humanKeywords: Information Systems research meth- interfaces implemented within the informationodologies, design science, design artifact, busi- system.ness environment, technology infrastructure,search strategies, experimental methods, The design-science paradigm has its roots increativity engineering and the sciences of the artificial (Simon 1996). It is fundamentally a problem- solving paradigm. It seeks to create innovations that define the ideas, practices, technical capa-Introduction bilities, and products through which the analysis, design, implementation, management, and use ofInformation systems are implemented within an information systems can be effectively andorganization for the purpose of improving the efficiently accomplished (Denning 1997;effectiveness and efficiency of that organization. Tsichritzis 1998). Such artifacts are not exemptCapabilities of the information system and char- from natural laws or behavioral theories. To theacteristics of the organization, its work systems, contrary, their creation relies on existing kernelits people, and its development and implemen- theories that are applied, tested, modified, andtation methodologies together determine the extended through the experience, creativity,extent to which that purpose is achieved (Silver et intuition, and problem solving capabilities of theal. 1995). It is incumbent upon researchers in the researcher (Markus et al. 2002; Walls et al. 1992).Information Systems (IS) discipline to “furtherknowledge that aids in the productive application The importance of design is well recognized in theof information technology to human organizations IS literature (Glass 1999; Winograd 1996, 1998).and their management” (ISR 2002, inside front Benbasat and Zmud (1999, p. 5) argue that thecover) and to develop and communicate “knowl- relevance of IS research is directly related to itsedge concerning both the management of applicability in design, stating that the implicationsinformation technology and the use of information of empirical IS research should be “implemen-technology for managerial and organizational pur- table,…synthesize an existing body of research,poses” (Zmud 1997). …[or] stimulate critical thinking” among IS practi- tioners. However, designing useful artifacts isWe argue that acquiring such knowledge involves complex due to the need for creative advances intwo complementary but distinct paradigms, domain areas in which existing theory is oftenbehavioral science and design science (March insufficient. “As technical knowledge grows, IT isand Smith 1995). The behavioral-science para- applied to new application areas that were notdigm has its roots in natural science research previously believed to be amenable to IT support”methods. It seeks to develop and justify theories (Markus et al. 2002, p. 180). The resultant IT(i.e., principles and laws) that explain or predict artifacts extend the boundaries of human problemorganizational and human phenomena sur- solving and organizational capabilities by pro-rounding the analysis, design, implementation, viding intellectual as well as computational tools.management, and use of information systems. Theories regarding their application and impactSuch theories ultimately inform researchers and will follow their development and use.practitioners of the interactions among people,technology, and organizations that must be Here, we argue, is an opportunity for IS researchmanaged if an information system is to achieve its to make significant contributions by engaging thestated purpose, namely improving the effective- complementary research cycle between design-76 MIS Quarterly Vol. 28 No. 1/March 2004
  3. 3. Hevner et al./Design Science in IS Researchscience and behavioral-science to address funda- artifact, the instantiation (system), although othermental problems faced in the productive applica- research efforts have also focused on thetion of information technology. Technology and evaluation of constructs (e.g., Batra et al. 1990;behavior are not dichotomous in an information Bodart et al. 2001; Geerts and McCarthy 2002;system. They are inseparable (Lee 2000). They Kim and March 1995) and methods (e.g., Marakasare similarly inseparable in IS research. Philo- and Elam 1998; Sinha and Vessey 1999).sophically these arguments draw from the prag- Relatively little behavioral research has focusedmatists (Aboulafia 1991) who argue that truth on evaluating models, a major focus of research(justified theory) and utility (artifacts that are in the management science literature.effective) are two sides of the same coin and thatscientific research should be evaluated in light of Design science, as the other side of the ISits practical implications. research cycle, creates and evaluates IT artifacts intended to solve identified organizational prob-The realm of IS research is at the confluence of lems. Such artifacts are represented in a struc-people, organizations, and technology (Davis and tured form that may vary from software, formalOlson 1985; Lee 1999). IT artifacts are broadly logic, and rigorous mathematics to informaldefined as constructs (vocabulary and symbols), natural language descriptions. A mathematicalmodels (abstractions and representations), basis for design allows many types of quantitativemethods (algorithms and practices), and instan- evaluations of an IT artifact, including optimizationtiations (implemented and prototype systems). proofs, analytical simulation, and quantitativeThese are concrete prescriptions that enable IT comparisons with alternative designs. The furtherresearchers and practitioners to understand and evaluation of a new artifact in a given organi-address the problems inherent in developing and zational context affords the opportunity to applysuccessfully implementing information systems empirical and qualitative methods. The richwithin organizations (March and Smith 1995; phenomena that emerge from the interaction ofNunamaker et al. 1991a). As illustrations, Markus people, organizations, and technology may needet al. (2002) and Walls et al. (1992) present to be qualitatively assessed to yield an under-design-science research aimed at developing standing of the phenomena adequate for theoryexecutive information systems (EISs) and systems development or problem solving (Klein andto support emerging knowledge processes Meyers 1999). As field studies enable behavioral-(EKPs), respectively, within the context of “IS science researchers to understand organizationaldesign theories.” Such theories prescribe “effec- phenomena in context, the process of constructingtive development practices” (methods) and “a type and exercising innovative IT artifacts enableof system solution” (instantiation) for “a particular design-science researchers to understand theclass of user requirements” (models) (Markus et problem addressed by the artifact and theal. 2002, p. 180). Such prescriptive theories must feasibility of their approach to its solutionbe evaluated with respect to the utility provided for (Nunamaker et al. 1991a).the class of problems addressed. The primary goal of this paper is to inform theAn IT artifact, implemented in an organizational community of IS researchers and practitioners ofcontext, is often the object of study in IS behav- how to conduct, evaluate, and present design-ioral-science research. Theories seek to predict science research. We do so by describing theor explain phenomena that occur with respect to boundaries of design science within the ISthe artifact’s use (intention to use), perceived discipline via a conceptual framework for under-usefulness, and impact on individuals and organi- standing information systems research and byzations (net benefits) depending on system, developing a set of guidelines for conducting andservice, and information quality (DeLone and evaluating good design-science research. WeMcLean 1992, 2003; Seddon 1997). Much of this focus primarily on technology-based designbehavioral research has focused on one class of although we note with interest the current explora- MIS Quarterly Vol. 28 No. 1/March 2004 77
  4. 4. Hevner et al./Design Science in IS Researchtion of organizations, policies, and work practices organizations to engage new forms and newas designed artifacts (Boland 2002). Following structures—to change the ways they “do busi-Klein and Myers (1999) treatise on the conduct ness” (Drucker 1988, 1991; Orlikowski 2000). Ourand evaluation of interpretive research in IS, we subsequent discussion of design science will beuse the proposed guidelines to assess recent limited to the activities of building the IS infrastruc-exemplar papers published in the IS literature in ture within the business organization. Issues oforder to illustrate how authors, reviewers, and strategy, alignment, and organizational infrastruc-editors can apply them consistently. We conclude ture design are outside the scope of this paper.with an analysis of the challenges of performinghigh-quality design-science research and a call for To achieve a true understanding of and appre-synergistic efforts between behavioral-science ciation for design science as an IS researchand design-science researchers. paradigm, an important dichotomy must be faced. Design is both a process (set of activities) and a product (artifact)—a verb and a noun (Walls et al. 1992). It describes the world as acted upon (pro-A Framework for IS Research cesses) and the world as sensed (artifacts). This Platonic view of design supports a problem-Information systems and the organizations they solving paradigm that continuously shifts perspec-support are complex, artificial, and purposefully tive between design processes and designeddesigned. They are composed of people, struc- artifacts for the same complex problem. Thetures, technologies, and work systems (Alter design process is a sequence of expert activities2003; Bunge 1985; Simon 1996). Much of the that produces an innovative product (i.e., thework performed by IS practitioners, and managers design artifact). The evaluation of the artifact thenin general (Boland 2002), deals with design—the provides feedback information and a betterpurposeful organization of resources to accom- understanding of the problem in order to improveplish a goal. Figure 1 illustrates the essential both the quality of the product and the designalignments between business and information process. This build-and-evaluate loop is typicallytechnology strategies and between organizational iterated a number of times before the final designand information systems infrastructures (Hender- artifact is generated (Markus et al. 2002). Duringson and Venkatraman 1993). The effective transi- this creative process, the design-science re-tion of strategy into infrastructure requires exten- searcher must be cognizant of evolving both thesive design activity on both sides of the figure— design process and the design artifact as part oforganizational design to create an effective the research.organizational infrastructure and informationsystems design to create an effective information March and Smith (1995) identify two designsystem infrastructure. processes and four design artifacts produced by design-science research in IS. The two processesThese are interdependent design activities that are build and evaluate. The artifacts are con-are central to the IS discipline. Hence, IS research structs, models, methods, and instantiations.must address the interplay among business Purposeful artifacts are built to address heretoforestrategy, IT strategy, organizational infrastructure, unsolved problems. They are evaluated withand IS infrastructure. This interplay is becoming respect to the utility provided in solving thosemore crucial as information technologies are seen problems. Constructs provide the language inas enablers of business strategy and organiza- which problems and solutions are defined andtional infrastructure (Kalakota and Robinson 2001; communicated (Schön 1983). Models use con-Orlikowski and Barley 2001). Available and structs to represent a real world situation—theemerging IT capabilities are a significant factor in design problem and its solution space (Simondetermining the strategies that guide an organiza- 1996). Models aid problem and solution under-tion. Cutting-edge information systems allow standing and frequently represent the connection78 MIS Quarterly Vol. 28 No. 1/March 2004
  5. 5. Hevner et al./Design Science in IS Research Strategy Information Business Alignment Technology Strategy Strategy Organizational Information Systems Design Activities Design Activities Infrastructure Organizational Alignment Information Infrastructure Systems Infrastructure Figure 1. Organizational Design and Information Systems Design Activities (Adapted from J. Henderson and N. Venkatraman, “Strategic Alignment: Leveraging Information Technology for Transforming Organizations,” IBM Systems Journal (32:1), 1993.)between problem and solution components people, (business) organizations, and theirenabling exploration of the effects of design existing or planned technologies (Silver et al.decisions and changes in the real world. Methods 1995). In it are the goals, tasks, problems, anddefine processes. They provide guidance on how opportunities that define business needs as theyto solve problems, that is, how to search the are perceived by people within the organization.solution space. These can range from formal, Such perceptions are shaped by the roles,mathematical algorithms that explicitly define the capabilities, and characteristics of people withinsearch process to informal, textual descriptions of the organization. Business needs are assessed“best practice” approaches, or some combination. and evaluated within the context of organizationalInstantiations show that constructs, models, or strategies, structure, culture, and existing busi-methods can be implemented in a working sys- ness processes. They are positioned relative totem. They demonstrate feasibility, enabling con- existing technology infrastructure, applications,crete assessment of an artifact’s suitability to its communication architectures, and developmentintended purpose. They also enable researchers capabilities. Together these define the businessto learn about the real world, how the artifact need or “problem” as perceived by the researcher.affects it, and how users appropriate it. Framing research activities to address business needs assures research relevance.Figure 2 presents our conceptual framework forunderstanding, executing, and evaluating IS Given such an articulated business need, ISresearch combining behavioral-science and research is conducted in two complementarydesign-science paradigms. We use this frame- phases. Behavioral science addresses researchwork to position and compare these paradigms. through the development and justification of theories that explain or predict phenomena relatedThe environment defines the problem space to the identified business need. Design science(Simon 1996) in which reside the phenomena of addresses research through the building andinterest. For IS research, it is composed of evaluation of artifacts designed to meet the iden- MIS Quarterly Vol. 28 No. 1/March 2004 79
  6. 6. Hevner et al./Design Science in IS Research Environment Relevance IS Research Rigor Knowledge Base People Foundations •Roles Develop/Build •Theories •Capabilities •Theories •Frameworks •Characteristics •Artifacts •Instruments •Constructs Organizations Business Applicable •Models •Strategies Needs Knowledge •Methods •Structure & Culture Assess Refine •Instantiations •Processes Methodologies Technology Justify/Evaluate •Data Analysis •Infrastructure •Analytical Techniques •Applications •Case Study •Formalisms •Communications •Experimental •Measures Architecture •Field Study •Validation Criteria •Development •Simulation Capabilities Application in the Additions to the Appropriate Environment Knowledge Base Figure 2. Information Systems Research Frameworktified business need. The goal of behavioral- artifact and the need to refine and reassess. Thescience research is truth.2 The goal of design- refinement and reassessment process is typicallyscience research is utility. As argued above, our described in future research directions.position is that truth and utility are inseparable.Truth informs design and utility informs theory. An The knowledge base provides the raw materialsartifact may have utility because of some as yet from and through which IS research is accom-undiscovered truth. A theory may yet to be devel- plished. The knowledge base is composed ofoped to the point where its truth can be incorpor- foundations and methodologies. Prior IS researchated into design. In both cases, research assess- and results from reference disciplines providement via the justify/evaluate activities can result in foundational theories, frameworks, instruments,the identification of weaknesses in the theory or constructs, models, methods, and instantiations used in the develop/build phase of a research study. Methodologies provide guidelines used in2 the justify/evaluate phase. Rigor is achieved by Theories posed in behavioral science are principled appropriately applying existing foundations andexplanations of phenomena. We recognize that suchtheories are approximations and are subject to numer- methodologies. In behavioral science, methodol-ous assumptions and conditions. However, they are ogies are typically rooted in data collection andevaluated against the norms of truth or explanatory empirical analysis techniques. In design science,power and are valued only as the claims they make areborne out in reality. computational and mathematical methods are80 MIS Quarterly Vol. 28 No. 1/March 2004
  7. 7. Hevner et al./Design Science in IS Researchprimarily used to evaluate the quality and effec- has produced a considerable literature on designtiveness of artifacts; however, empirical techni- (Dym 1994; Pahl and Beitz 1996; Petroski 1996).ques may also be employed. Within the IS discipline, many design activities have been extensively studied, formalized, andThe contributions of behavioral science and become normal or routine. Design-sciencedesign science in IS research are assessed as research in IS addresses what are considered tothey are applied to the business need in an be wicked problems (Brooks 1987, 1996; Rittelappropriate environment and as they add to the and Webber 1984). That is, those problemscontent of the knowledge base for further research characterized byand practice. A justified theory that is not usefulfor the environment contributes as little to the IS • unstable requirements and constraints basedliterature as an artifact that solves a nonexistent upon ill-defined environmental contextsproblem. • complex interactions among subcomponentsOne issue that must be addressed in design- of the problem and its solutionscience research is differentiating routine designor system building from design research. The • inherent flexibility to change design pro-difference is in the nature of the problems andsolutions. Routine design is the application of cesses as well as design artifacts (i.e.,existing knowledge to organizational problems, malleable processes and artifacts)such as constructing a financial or marketinginformation system using best practice artifacts • a critical dependence upon human cognitive(constructs, models, methods, and instantiations) abilities (e.g., creativity) to produce effectiveexisting in the knowledge base. On the other solutionshand, design-science research addresses impor-tant unsolved problems in unique or innovative • a critical dependence upon human socialways or solved problems in more effective or abilities (e.g., teamwork) to produce effectiveefficient ways. The key differentiator between rou- solutionstine design and design research is the clear iden-tification of a contribution to the archival knowl-edge base of foundations and methodologies. As a result, we agree with Simon (1996) that a theory of design in information systems, ofIn the early stages of a discipline or with signifi- necessity, is in a constant state of scientificcant changes in the environment, each new revolution (Kuhn 1996). Technological advancesartifact created for that discipline or environment are the result of innovative, creative designis “an experiment” that “poses a question to science processes. If not capricious, they are atnature” (Newell and Simon 1976, p 114). Existing least arbitrary (Brooks 1987) with respect toknowledge is used where appropriate; however, business needs and existing knowledge.often the requisite knowledge is nonexistent Innovations, such as database management sys-(Markus et al. 2002). Reliance on creativity and tems, high-level languages, personal computers,trial-and-error search are characteristic of such software components, intelligent agents, objectresearch efforts. As design-science researchresults are codified in the knowledge base, they technology, the Internet, and the World Widebecome best practice. System building is then the Web, have had dramatic and at times unintendedroutine application of the knowledge base to impacts on the way in which information systemsknown problems. are conceived, designed, implemented, and managed. Consequently the guidelines weDesign activities are endemic in many profes- present below are, of necessity, adaptive andsions. In particular, the engineering profession process-oriented. MIS Quarterly Vol. 28 No. 1/March 2004 81
  8. 8. Hevner et al./Design Science in IS ResearchGuidelines for Design Science contend that each of these guidelines should be addressed in some manner for design-sciencein Information Systems research to be complete. How well the researchResearch satisfies the intent of each of the guidelines is then a matter for the reviewers, editors, andAs discussed above, design science is inherently readers to determine.a problem solving process. The fundamentalprinciple of design-science research from which Table 1 summarizes the seven guidelines. Eachour seven guidelines are derived is that knowl- is discussed in detail below. In the followingedge and understanding of a design problem and section, they are applied to specific exemplarits solution are acquired in the building and research efforts.application of an artifact. That is, design-scienceresearch requires the creation of an innovative,purposeful artifact (Guideline 1) for a specifiedproblem domain (Guideline 2). Because the Guideline 1: Design as an Artifactartifact is purposeful, it must yield utility for thespecified problem. Hence, thorough evaluation of The result of design-science research in IS is, bythe artifact is crucial (Guideline 3). Novelty is definition, a purposeful IT artifact created to ad-similarly crucial since the artifact must be dress an important organizational problem. Itinnovative, solving a heretofore unsolved problem must be described effectively, enabling its imple-or solving a known problem in a more effective or mentation and application in an appropriateefficient manner (Guideline 4). In this way, domain.design-science research is differentiated from thepractice of design. The artifact itself must be Orlikowski and Iacono (2001) call the IT artifactrigorously defined, formally represented, coherent, the “core subject matter” of the IS field. Althoughand internally consistent (Guideline 5). The pro- they articulate multiple definitions of the term ITcess by which it is created, and often the artifact artifact, many of which include components of theitself, incorporates or enables a search process organization and people involved in the use of awhereby a problem space is constructed and a computer-based artifact, they emphasize themechanism posed or enacted to find an effective importance of “those bundles of cultural propertiessolution (Guideline 6). Finally, the results of the packaged in some socially recognizable form suchdesign-science research must be communicated as hardware and software” (p. 121), i.e., the ITeffectively (Guideline 7) both to a technical artifact as an instantiation. Weber (1987) arguesaudience (researchers who will extend them and that theories of long-lived artifacts (instantiations)practitioners who will implement them) and to a and their representations (Weber 2003) aremanagerial audience (researchers who will study fundamental to the IS discipline. Such theoriesthem in context and practitioners who will decide must explain how artifacts are created andif they should be implemented within their adapted to their changing environments andorganizations). underlying technologies.Our purpose for establishing these seven Our definition of IT artifacts is both broader andguidelines is to assist researchers, reviewers, narrower then those articulated above. It iseditors, and readers to understand the require- broader in the sense that we include not onlyments for effective design-science research. instantiations in our definition of the IT artifact butFollowing Klein and Myers (1999), we advise also the constructs, models, and methods appliedagainst mandatory or rote use of the guidelines. in the development and use of informationResearchers, reviewers, and editors must use systems. However, it is narrower in the sense thattheir creative skills and judgment to determine we do not include people or elements of organi-when, where, and how to apply each of the guide- zations in our definition nor do we explicitlylines in a specific research project. However, we include the process by which such artifacts evolve82 MIS Quarterly Vol. 28 No. 1/March 2004
  9. 9. Hevner et al./Design Science in IS Research Table 1. Design-Science Research Guidelines Guideline Description Guideline 1: Design as an Artifact Design-science research must produce a viable artifact in the form of a construct, a model, a method, or an instantiation. Guideline 2: Problem Relevance The objective of design-science research is to develop technology-based solutions to important and relevant business problems. Guideline 3: Design Evaluation The utility, quality, and efficacy of a design artifact must be rigorously demonstrated via well-executed evaluation methods. Guideline 4: Research Contributions Effective design-science research must provide clear and verifiable contributions in the areas of the design artifact, design foundations, and/or design methodologies. Guideline 5: Research Rigor Design-science research relies upon the application of rigorous methods in both the construction and evaluation of the design artifact. Guideline 6: Design as a Search The search for an effective artifact requires utilizing available Process means to reach desired ends while satisfying laws in the problem environment. Guideline 7: Communication of Design-science research must be presented effectively both Research to technology-oriented as well as management-oriented audiences.over time. We conceive of IT artifacts not as 1997; Tsichritzis 1998). This definition of theindependent of people or the organizational and artifact is consistent with the concept of IS designsocial contexts in which they are used but as theory as used by Walls et al. (1992) and Markusinterdependent and coequal with them in meeting et al. (2002) where the theory addresses both thebusiness needs. We acknowledge that percep- process of design and the designed product.tions and fit with an organization are crucial to thesuccessful development and implementation of an More precisely, constructs provide the vocabularyinformation system. We argue, however, that the and symbols used to define problems andcapabilities of the constructs, models, methods, solutions. They have a significant impact on theand instantiations are equally crucial and that way in which tasks and problems are conceiveddesign-science research efforts are necessary for (Boland 2002; Schön 1983). They enable thetheir creation. construction of models or representations of the problem domain. Representation has a profound impact on design work. The field of mathematicsFurthermore, artifacts constructed in design- was revolutionized, for example, with the con-science research are rarely full-grown information structs defined by Arabic numbers, zero, andsystems that are used in practice. Instead, artif- place notation. The search for an effective prob-acts are innovations that define the ideas, lem representation is crucial to finding an effectivepractices, technical capabilities, and products design solution (Weber 2003). Simon (1996, p.through which the analysis, design, implemen- 132) states, “solving a problem simply meanstation, and use of information systems can be representing it so as to make the solutioneffectively and efficiently accomplished (Denning transparent.” MIS Quarterly Vol. 28 No. 1/March 2004 83
  10. 10. Hevner et al./Design Science in IS ResearchThe entity-relationship model (Chen 1976), for To illustrate further, prior to the construction of theexample, is a set of constructs for representing first expert system (instantiation), it was not clearthe semantics of data. It has had a profound if such a system could be constructed. It was notimpact on the way in which systems analysis and clear how to describe or represent it, or how welldatabase design are executed and the way in it would perform. Once feasibility was demon-which information systems are represented and strated by constructing an expert system in adeveloped. Furthermore, these constructs have selected domain, constructs and models werebeen used to build models of specific business developed and subsequent research in expert systems focused on demonstrating significantsituations that have been generalized into patterns improvements in the product or process (methods)for application in similar domains (Purao et al. of construction (Tam 1990; Trice and Davis 1993).2003). Methods for building such models have Similar examples exist in requirements determi-also been the subject of considerable research nation (Bell 1993; Bhargava et al. 1998), individual(Halpin 2001; McCarthy 1982; Parsons and Wand and group decision support systems (Aiken et al.2000; Storey et al. 1997). 1991; Basu and Blanning 1994), database design and integration (Dey et al. 1998; Dey et al. 1999;Artifact instantiation demonstrates feasibility both Storey et al. 1997), and workflow analysis (Basuof the design process and of the designed pro- and Blanning 2000), to name a few importantduct. Design-science research in IT often ad- areas of IS design-science research.dresses problems related to some aspect of thedesign of an information system. Hence, theinstantiations produced may be in the form of Guideline 2: Problem Relevanceintellectual or software tools aimed at improvingthe process of information system development. The objective of research in information systemsConstructing a system instantiation that auto- is to acquire knowledge and understanding thatmates a process demonstrates that the process enable the development and implementation ofcan, in fact, be automated. It provides “proof by technology-based solutions to heretofore unsolvedconstruction” (Nunamaker 1991a). The critical and important business problems. Behavioralnature of design-science research in IS lies in the science approaches this goal through the devel-identification of as yet undeveloped capabilities opment and justification of theories explaining orneeded to expand IS into new realms “not predicting phenomena that occur. Design sciencepreviously believed amenable to IT support” approaches this goal through the construction of(Markus et al. 2002, p. 180). Such a result is innovative artifacts aimed at changing the pheno-significant IS research only if there is a serious mena that occur. Each must inform and challenge the other. For example, the technology accep-question about the ability to construct such an tance model provides a theory that explains andartifact, there is uncertainty about its ability to predicts the acceptance of information techno-perform appropriately, and the automated task is logies within organizations (Venkatesh 2000).important to the IS community. TOP Modeler This theory challenges design-science re-(Markus et al. 2002), for example, is a tool that searchers to create artifacts that enable organi-instantiates methods for the development of zations to overcome the acceptance problemsinformation systems that support “emergent predicted. We argue that a combination ofknowledge processes.” Construction of such a technology-based artifacts (e.g., system concep-prototype artifact in a research setting or in a tualizations and representations, practices, tech-single organizational setting is only a first step nical capabilities, interfaces, etc.), organization-toward its deployment, but we argue that it is a based artifacts (e.g., structures, compensation,necessary one. As an exemplar of design-science reporting relationships, social systems, etc.), andresearch (see below), this research resulted in a people-based artifacts (e.g., training, consensuscommercial product that “has been used in over building, etc.) are necessary to address suchtwo dozen ‘real use’ situations” (p. 187). issues.84 MIS Quarterly Vol. 28 No. 1/March 2004
  11. 11. Hevner et al./Design Science in IS ResearchFormally, a problem can be defined as the crucial component of the research process. Thedifferences between a goal state and the current business environment establishes the require-state of a system. Problem solving can be defined ments upon which the evaluation of the artifact isas a search process (see Guideline 6) using based. This environment includes the technicalactions to reduce or eliminate the differences infrastructure which itself is incrementally built by(Simon 1996). These definitions imply an environ- the implementation of new IT artifacts. Thus,ment that imposes goal criteria as well as evaluation includes the integration of the artifactconstraints upon a system. Business organiza- within the technical infrastructure of the businesstions are goal-oriented entities existing in an environment.economic and social setting. Economic theoryoften portrays the goals of business organizations As in the justification of a behavioral scienceas being related to profit (utility) maximization. theory, evaluation of a designed IT artifactHence, business problems and opportunities often requires the definition of appropriate metrics andrelate to increasing revenue or decreasing cost possibly the gathering and analysis of appropriatethrough the design of effective business pro- data. IT artifacts can be evaluated in terms ofcesses. The design of organizational and inter- functionality, completeness, consistency, accu-organizational information systems plays a major racy, performance, reliability, usability, fit with therole in enabling effective business processes to organization, and other relevant quality attributes.achieve these goals. When analytical metrics are appropriate, designed artifacts may be mathematically evaluated. AsThe relevance of any design-science research two examples, distributed database design algo-effort is with respect to a constituent community. rithms can be evaluated using expected operatingFor IS researchers, that constituent community is cost or average response time for a giventhe practitioners who plan, manage, design, characterization of information processing require-implement, operate, and evaluate information ments (Johansson et al. 2003) and searchsystems and those who plan, manage, design, algorithms can be evaluated using informationimplement, operate, and evaluate the tech- retrieval metrics such as precision and recallnologies that enable their development and (Salton 1988).implementation. To be relevant to this community,research must address the problems faced and Because design is inherently an iterative andthe opportunities afforded by the interaction of incremental activity, the evaluation phase providespeople, organizations, and information technology. essential feedback to the construction phase as toOrganizations spend billions of dollars annually on the quality of the design process and the designIT, only too often to conclude that those dollars product under development. A design artifact iswere wasted (Keil 1995; Keil et al. 1998; Keil and complete and effective when it satisfies theRobey 1999). This community would welcome requirements and constraints of the problem iteffective artifacts that enable such problems to be was meant to solve. Design-science researchaddressed—constructs by which to think about efforts may begin with simplified conceptuali-them, models by which to represent and explore zations and representations of problems. Asthem, methods by which to analyze or optimize available technology or organizational environ-them, and instantiations that demonstrate how to ments change, assumptions made in prioraffect them. research may become invalid. Johansson (2000), for example, demonstrated that network latency is a major component in the response-time perfor-Guideline 3: Design Evaluation mance of distributed databases. Prior research in distributed database design ignored latencyThe utility, quality, and efficacy of a design artifact because it assumed a low-bandwidth networkmust be rigorously demonstrated via well- where latency is negligible. In a high-bandwidthexecuted evaluation methods. Evaluation is a network, however, latency can account for over 90 MIS Quarterly Vol. 28 No. 1/March 2004 85
  12. 12. Hevner et al./Design Science in IS Research Table 2. Design Evaluation Methods 1. Observational Case Study: Study artifact in depth in business environment Field Study: Monitor use of artifact in multiple projects 2. Analytical Static Analysis: Examine structure of artifact for static qualities (e.g., complexity) Architecture Analysis: Study fit of artifact into technical IS architecture Optimization: Demonstrate inherent optimal properties of artifact or provide optimality bounds on artifact behavior Dynamic Analysis: Study artifact in use for dynamic qualities (e.g., performance) 3. Experimental Controlled Experiment: Study artifact in controlled environment for qualities (e.g., usability) Simulation – Execute artifact with artificial data 4. Testing Functional (Black Box) Testing: Execute artifact interfaces to discover failures and identify defects Structural (White Box) Testing: Perform coverage testing of some metric (e.g., execution paths) in the artifact implementation 5. Descriptive Informed Argument: Use information from the knowledge base (e.g., relevant research) to build a convincing argument for the artifact’s utility Scenarios: Construct detailed scenarios around the artifact to demonstrate its utilitypercent of the response time. Johansson et al. Design, in all of its realizations (e.g., architecture,(2003) extended prior distributed database design landscaping, art, music), has style. Given theresearch by developing a model that includes problem and solution requirements, sufficientnetwork latency and the effects of parallel pro- degrees of freedom remain to express a variety ofcessing on response time. forms and functions in the artifact that are aesthetically pleasing to both the designer and theThe evaluation of designed artifacts typically uses user. Good designers bring an element of style tomethodologies available in the knowledge base. their work (Norman 1988). Thus, we posit thatThese are summarized in Table 2. The selection design evaluation should include an assessmentof evaluation methods must be matched appro- of the artifact’s style.priately with the designed artifact and the selectedevaluation metrics. For example, descriptive The measurement of style lies in the realm ofmethods of evaluation should only be used for human perception and taste. In other words, weespecially innovative artifacts for which other know good style when we see it. While difficult toforms of evaluation may not be feasible. The define, style in IS design is widely recognized andgoodness and efficacy of an artifact can be appreciated (Kernighan and Plauger 1978; Wino-rigorously demonstrated via well-selected evalua- grad 1996). Gelernter (1998) terms the essencetion methods (Basili 1996; Kleindorfer et al. 1998; of style in IS design machine beauty. He de-Zelkowitz and Wallace 1998). scribes it as a marriage between simplicity and86 MIS Quarterly Vol. 28 No. 1/March 2004
  13. 13. Hevner et al./Design Science in IS Researchpower that drives innovation in science and formalisms, ontologies (Wand and Webertechnology. Simon (1996) also notes the impor- 1993, 1995; Weber 1997), problem andtance of style in the design process. The ability to solution representations, design algorithmscreatively vary the design process, within the (Storey et al. 1997), and innovativelimits of satisfactory constraints, challenges and information systems (Aiken 1991; Markus etadds value to designers who participate in the al. 2002; Walls et al. 1992) are examples ofprocess. such artifacts. 3. Methodologies. Finally, the creative develop-Guideline 4: Research Contributions ment and use of evaluation methods (e.g., experimental, analytical, observational,Effective design-science research must provide testing, and descriptive) and new evaluationclear contributions in the areas of the design metrics provide design-science researchartifact, design construction knowledge (i.e., foun- contributions. Measures and evaluationdations), and/or design evaluation knowledge (i.e., metrics in particular are crucial componentsmethodologies). The ultimate assessment for any of design-science research. The right-facingresearch is, “What are the new and interesting arrow at the bottom of the figure from IScontributions?” Design-science research holds Research to the Knowledge Base in Figure 2the potential for three types of research contri- also indicates these contributions. TAM, forbutions based on the novelty, generality, and example, presents a framework for predictingsignificance of the designed artifact. One or more and explaining why a particular informationof these contributions must be found in a given system will or will not be accepted in a givenresearch project. organizational setting (Venkatesh 2000). Although TAM is posed as a behavioral1. The Design Artifact. Most often, the contribu- theory, it also provides metrics by which a tion of design-science research is the artifact designed information system or implemen- itself. The artifact must enable the solution of tation process can be evaluated. Its implica- heretofore unsolved problems. It may extend tions for design itself are as yet unexplored. the knowledge base (see below) or apply existing knowledge in new and innovative Criteria for assessing contribution focus on ways. As shown in Figure 2 by the left-facing representational fidelity and implementability. arrow at the bottom of the figure from IS Artifacts must accurately represent the business Research to the Environment, exercising the and technology environments used in the artifact in the environment produces research, information systems themselves being significant value to the constituent IS models of the business. These artifacts must be community. System development method- “implementable,” hence the importance of instan- ologies, design tools, and prototype systems tiating design science artifacts. Beyond these, (e.g., GDSS, expert systems) are examples however, the research must demonstrate a clear of such artifacts. contribution to the business environment, solving an important, previously unsolved problem.2. Foundations. The creative development of novel, appropriately evaluated constructs, models, methods, or instantiations that Guideline 5: Research Rigor extend and improve the existing foundations in the design-science knowledge base are Rigor addresses the way in which research is also important contributions. The right-facing conducted. Design-science research requires the arrow at the bottom of the figure from IS application of rigorous methods in both the Research to the Knowledge Base in Figure 2 construction and evaluation of the designed indicates these contributions. Modeling artifact. In behavioral-science research, rigor is MIS Quarterly Vol. 28 No. 1/March 2004 87
  14. 14. Hevner et al./Design Science in IS Researchoften assessed by adherence to appropriate data comparability, subject selection, training, time,collection and analysis techniques. Over- and tasks. Methods for this type of evaluation areemphasis on rigor in behavioral IS research has not unlike those for justifying or testing behavioraloften resulted in a corresponding lowering of theories. However, the principal aim is to deter-relevance (Lee 1999). mine how well an artifact works, not to theorize about or prove anything about why the artifactDesign-science research often relies on mathe- works. This is where design-science andmatical formalism to describe the specified and behavioral-science researchers must complementconstructed artifact. However, the environments one another. Because design-science artifactsin which IT artifacts must perform and the artifacts are often the “machine” part of the human-themselves may defy excessive formalism. Or, in machine system constituting an information sys-an attempt to be mathematically rigorous, tem, it is imperative to understand why an artifactimportant parts of the problem may be abstracted works or does not work to enable new artifacts toor “assumed away.” In particular, with respect to be constructed that exploit the former and avoidthe construction activity, rigor must be assessed the latter.with respect to the applicability and generali-zability of the artifact. Again, an overemphasis onrigor can lessen relevance. We argue, along with Guideline 6: Design as abehavioral IS researchers (Applegate 1999), that Search Processit is possible and necessary for all IS researchparadigms to be both rigorous and relevant. Design science is inherently iterative. The search for the best, or optimal, design is often intractableIn both design-science and behavioral-science for realistic information systems problems.research, rigor is derived from the effective use of Heuristic search strategies produce feasible, goodthe knowledge base—theoretical foundations and designs that can be implemented in the businessresearch methodologies. Success is predicated environment. Simon (1996) describes the natureon the researcher’s skilled selection of appropriate of the design process as a Generate/Test Cycletechniques to develop or construct a theory or (Figure 3).artifact and the selection of appropriate means tojustify the theory or evaluate the artifact. Design is essentially a search process to discover an effective solution to a problem. ProblemClaims about artifacts are typically dependent solving can be viewed as utilizing available meansupon performance metrics. Even formal mathe- to reach desired ends while satisfying lawsmatical proofs rely on evaluation criteria against existing in the environment (Simon 1996).which the performance of an artifact can be Abstraction and representation of appropriatemeasured. Design-science researchers must means, ends, and laws are crucial components ofconstantly assess the appropriateness of their design-science research. These factors are prob-metrics and the construction of effective metrics is lem and environment dependent and invariablyan important part of design-science research. involve creativity and innovation. Means are the set of actions and resources available to constructFurthermore, designed artifacts are often com- a solution. Ends represent goals and constraintsponents of a human-machine problem-solving on the solution. Laws are uncontrollable forces insystem. For such artifacts, knowledge of behav- the environment. Effective design requires knowl-ioral theories and empirical work are necessary to edge of both the application domain (e.g., require-construct and evaluate such artifacts. Constructs, ments and constraints) and the solution domainmodels, methods, and instantiations must be (e.g., technical and organizational).exercised within appropriate environments.Appropriate subject groups must be obtained for Design-science research often simplifies a prob-such studies. Issues that are addressed include lem by explicitly representing only a subset of the88 MIS Quarterly Vol. 28 No. 1/March 2004
  15. 15. Hevner et al./Design Science in IS Research Generate Design Alternatives Test Alternatives Against Requirements/Constraints Figure 3. The Generate/Test Cyclerelevant means, ends, and laws or by decom- may not be possible to determine, let aloneposing a problem into simpler subproblems. Such explicitly describe, the relevant means, ends, orsimplifications and decompositions may not be laws (Vessey and Glass 1998). Even when it isrealistic enough to have a significant impact on possible to do so, the sheer size and complexity ofpractice but may represent a starting point. the solution space will often render the problemProgress is made iteratively as the scope of the computationally infeasible. For example, to builddesign problem is expanded. As means, ends, a “reliable, secure, and responsive informationand laws are refined and made more realistic, the systems infrastructure,” one of the key issuesdesign artifact becomes more relevant and faced by IS managers (Brancheau et al. 1996), avaluable. The means, ends, and laws for IS designer would need to represent all possibledesign problems can often be represented using infrastructures (means), determine their utility andthe tools of mathematics and operations research. constraints (ends), and specify all cost and benefitMeans are represented by decision variables constants (laws). Clearly such an approach iswhose values constitute an implementable design infeasible. However, this does not mean thatsolution. Ends are represented using a utility design-science research is inappropriate for suchfunction and constraints that can be expressed in a problem.terms of decision variables and constants. Lawsare represented by the values of constants used In such situations, the search is for satisfactoryin the utility function and constraints. solutions, i.e., satisficing (Simon 1996), without explicitly specifying all possible solutions. TheThe set of possible design solutions for any design task involves the creation, utilization, andproblem is specified as all possible means that assessment of heuristic search strategies. Thatsatisfy all end conditions consistent with identified is, constructing an artifact that “works” well for thelaws. When these can be formulated appro- specified class of problems. Although its con-priately and posed mathematically, standard struction is based on prior theory and existingoperations research techniques can be used to design knowledge, it may or may not be entirelydetermine an optimal solution for the specified clear why it works or the extent of its generaliza-end conditions. Given the wicked nature of many bility; it simply qualifies as “credentialed knowl-information system design problems, however, it edge” (Meehl 1986, p. 311). While it is important MIS Quarterly Vol. 28 No. 1/March 2004 89
  16. 16. Hevner et al./Design Science in IS Researchto understand why an artifact works, the critical should be committed to constructing (or pur-nature of design in IS makes it important to first chasing) and using the artifact within their specificestablish that it does work and to characterize the organizational context. Zmud (1997) suggestsenvironments in which it works, even if we cannot that presentation of design-science research for acompletely explain why it works. This enables IS managerial audience requires an emphasis not onpractitioners to take advantage of the artifact to the inherent nature of the artifact itself, but on theimprove practice and provides a context for knowledge required to effectively apply the artifactadditional research aimed at more fully explicating “within specific contexts for individual or organi-the resultant phenomena. Markus et al. (2002), zational gain” (p. ix). That is, the emphasis mustfor example, describe their search process in be on the importance of the problem and theterms of iteratively identifying deficiencies in novelty and effectiveness of the solution approachconstructed prototype software systems and realized in the artifact. While we agree with thiscreatively developing solutions to address them. statement, we note that it may be necessary to describe the artifact in some detail to enableThe use of heuristics to find “good” design solu- managers to appreciate its nature and understandtions opens the question of how goodness is its application. Presenting that detail in concise,measured. Different problem representations may well-organized appendices, as advised by Zmud,provide varying techniques for measuring how is an appropriate communication mechanism forgood a solution is. One approach is to prove or such an audience.demonstrate that a heuristic design solution isalways within close proximity of an optimal solu-tion. Another is to compare produced solutionswith those constructed by expert human designersfor the same problem situation. Application of the Design Science Research GuidelinesGuideline 7: Communicationof Research To illustrate the application of the design-science guidelines to IS research, we have selected threeDesign-science research must be presented both exemplar articles for analysis from three differentto technology-oriented as well as management- IS journals, one from Decision Support Systems,oriented audiences. Technology-oriented audi- one from Information Systems Research, and oneences need sufficient detail to enable the from MIS Quarterly. Each has strengths anddescribed artifact to be constructed (implemented) weaknesses when viewed through the lens of theand used within an appropriate organizational above guidelines. Our goal is not to perform acontext. This enables practitioners to take advan- critical evaluation of the quality of the researchtage of the benefits offered by the artifact and it contributions, but rather to illuminate the design-enables researchers to build a cumulative knowl- science guidelines. The articles areedge base for further extension and evaluation. Itis also important for such audiences to under- • Gavish and Gerdes (1998), which developsstand the processes by which the artifact was techniques for implementing anonymity inconstructed and evaluated. This establishes Group Decision Support Systems (GDSS)repeatability of the research project and builds the environmentsknowledge base for further research extensions bydesign-science researchers in IS. • Aalst and Kumar (2003), which proposes a design for an eXchangeable Routing Lan-Management-oriented audiences need sufficient guage (XRL) to support electronic commercedetail to determine if the organizational resources workflows among trading partners90 MIS Quarterly Vol. 28 No. 1/March 2004
  17. 17. Hevner et al./Design Science in IS Research• Markus, Majchrzak, and Gasser (2002), GDSS environment and then study the individual, which proposes a design theory for the group, or organizational implications using a development of information systems built to behavioral-science research paradigm. Several support emergent knowledge processes such GDSS papers have appeared in MIS Quarterly (e.g., Dickson et al. 1993; Gallupe et al.The fundamental questions for design-science 1988; Jarvenpaa et al. 1988; Sengupta and Te’eniresearch are, “What utility does the new artifact 1993).provide?” and “What demonstrates that utility?”Evidence must be presented to address these two The central role of design science in GDSS isquestions. That is the essence of design science. clearly recognized in the early foundation papersContribution arises from utility. If existing artifacts of the field. The University of Arizona Electronicare adequate, then design-science research that Meeting System group, for example, states thecreates a new artifact is unnecessary (it is need for both developmental and empiricalirrelevant). If the new artifact does not map ade- research agendas (Dennis et al. 1988; Nuna-quately to the real world (rigor), it cannot provide maker et al. 1991b). Developmental, or design-utility. If the artifact does not solve the problem science, research is called for in the areas of(search, implementability), it has no utility. If utility process structures and support and task struc-is not demonstrated (evaluation), then there is no tures and support. Process structure and supportbasis upon which to accept the claims that it technologies and methods are generic to allprovides any contribution (contribution). Further- GDSS environments and tasks. Technologiesmore, if the problem, the artifact, and its utility are and methods for distributed communications,not presented in a manner such that the implica- group memory, decision-making methods, andtions for research and practice are clear, then anonymity are a few of the critical design issuespublication in the IS literature is not appropriate for GDSS process support needed in any task(communication). domain. Task structure and support are specific to the problem domain under consideration by the group (e.g., medical decision making, softwareThe Design and Implementation development). Task support includes the designof Anonymity in GDSS: of new technologies and methods for managingGavish and Gerdes and analyzing task-related information and using that information to make specific, task-relatedThe study of group decision support systems decisions.(GDSS) has been and remains one of the mostvisible and successful research streams in the IS The issue of anonymity has been studiedfield. The use of information technology to effec- extensively in GDSS environments. Behavioraltively support meetings of groups of different sizes research studies have shown both positive andover time and space is a real problem that negative impacts on group interactions. On thechallenges all business organizations. Recent positive side, GDSS participants can express theirGDSS literature surveys demonstrate the large views freely without fear of embarrassment ornumbers of GDSS research papers published in reprisal. However, anonymity can encourage free-the IS field and, more importantly, the wide variety riding and antisocial behaviors. While the prosof research paradigms applied to GDSS research and cons of anonymity in GDSS are much(e.g., Dennis and Wixom 2001; Fjermestad and researched, there has been a noticeable lack ofHiltz 1998; Nunamaker et al. 1996). However, research on the design of techniques for imple-only a small number of GDSS papers can be menting anonymity in GDSS environments.considered to make true design-science research Gavish and Gerdes (1998) address this issue bycontributions. Most assume the introduction of a designing five basic mechanisms to providenew information technology or process in the GDSS procedural anonymity. MIS Quarterly Vol. 28 No. 1/March 2004 91
  18. 18. Hevner et al./Design Science in IS ResearchProblem Relevance design mechanisms to satisfy the system requirements for anonymity. Proposed designsThe amount of interest and research on anonymity are presented and anonymity claims are proved toissues in GDSS testifies to its relevance. Field be correct. A thorough discussion of the costsstudies and surveys clearly indicate that partici- and benefits of the proposed anonymitypants rank anonymity as a highly desired attribute mechanisms is provided in Section 4 of the paper.in the GDSS system. Many individuals state thatthey would refuse to participate in or trust theresults of a GDSS meeting without a satisfactory Design as an Artifactlevel of assured anonymity (Fjermestad and Hiltz1998). The authors design a GDSS system architecture that provides a rigorous level of procedural anonymity. Five mechanisms are employed toResearch Rigor ensure participant anonymity:Gavish and Gerdes base their GDSS anonymity • All messages are encrypted with a uniquedesigns on past research in the fields of crypto- session keygraphy and secure network communication proto-cols (e.g., Chaum 1981; Schneier 1996). These • The sender’s header information is removedresearch areas have a long history of formal, from all messagesrigorous results that have been applied to thedesign of many practical security and privacy • All messages are re-encrypted upon retrans-mechanisms. Appendix A of the exemplar paper mission from any GDSS serverprovides a set of formal proofs that the claimsmade by the authors for the anonymity designs • Transmission order of messages is ran-are correct and draw their validity from the domizedknowledge base of this past research. • Artificial messages are introduced to thwart traffic analysisDesign as a Search Process The procedures and communication protocols thatThe authors motivate their design science implement these mechanisms in a GDSS systemresearch by identifying three basic types of anony- are the artifacts of this research.mity in a GDSS system: environmental, content,and procedural. After a definition and brief dis-cussion of each type, they focus on the design of Design Evaluationmechanisms for procedural anonymity; the abilityof the GDSS system to hide the source of any The evaluation consists of two reported activities.message. This is a very difficult requirement First, in Appendix A, each mechanism is proved tobecause standard network protocols typically correctly provide the claimed anonymity benefits.attach source information in headers to support Formal proof methods are used to validate thereliable transmission protocols. Thus, GDSS sys- effectiveness of the designed mechanisms.tems must modify standard communication proto- Second, Section 4 presents a thorough cost-cols and include additional transmission proce- benefit analysis. It is shown that the operationaldures to ensure required levels of anonymity. costs of supporting the proposed anonymity mechanisms can be quite significant. In addition,The design-science process employed by the the communication protocols to implement theauthors is to state the desired procedural anony- mechanisms add considerable complexity to themity attributes of the GDSS system and then to system. Thus, the authors recommend that a92 MIS Quarterly Vol. 28 No. 1/March 2004
  19. 19. Hevner et al./Design Science in IS Researchcost-benefit justification be performed before business processes. Workflow managementdetermining the level of anonymity to implement systems are becoming integral components offor a GDSS meeting. many commercial enterprise-wide information systems (Leymann and Roller 2000). StandardsThe authors do not claim to have implemented the for workflow semantics and syntax (i.e., workflowproposed anonymity mechanisms in a prototype languages) and workflow architectures areor actual GDSS system. Thus, an instantiation of promulgated by the Workflow Managementthe designed artifact remains to be evaluated in Coalition (WfMC 2000). While workflow modelsan operational GDSS environment. have been used for many years to manage intra- organizational business processes, there is now a great demand for effective tools to model inter-Research Contributions organization processes across heterogeneous and distributed environments, such as those foundThe design-science contributions of this research in electronic commerce and complex supplyare the proposed anonymity mechanisms as the chains (Kumar and Zhao 2002).design artifacts and the evaluation results in theform of formal proofs and cost-benefit analyses. Aalst and Kumar (2003) investigate the problem ofThese contributions advance our understanding of exchanging business process information acrosshow best to provide participant anonymity inGDSS meetings. multiple organizations in an automated manner. They design an eXchangable Routing Language (XRL) to capture workflow models that are thenResearch Communication embedded in eXtensible Markup Language (XML) for electronic transmission to all participants in anAlthough the presentation of this research is interorganizational business process. The designaimed at an audience familiar with network system of XRL is based upon Petri nets, which provide aconcepts such as encryption and communication formal basis for analyzing the correctness andprotocols, the paper also contains important, performance of the workflows, as well asuseful information for a managerial audience. supporting the extensibility of the language. TheManagers should have a good understanding of authors develop a workflow management archi-the implications of anonymity in GDSS meetings. tecture and a prototype implementation toThis understanding must include an appreciation evaluate XRL in a proof of concept.of the costs of providing desired levels ofparticipant anonymity. While the authors providea thorough discussion of cost-benefit tradeoffs Problem Relevancetoward the end of the paper, the paper would bemore accessible to a managerial audience if it Interorganizational electronic commerce isincluded a stronger motivation up front on the growing rapidly and is projected to soon exceedimportant implications of anonymity in GDSS one trillion dollars annually (eMarketer 2002). Asystem development and operations. multitude of electronic commerce solutions are being proposed (e.g., ebXML, UDDI, RosettaNet) to enable businesses to execute transactions inA Workflow Language for Inter- standardized, open environments. While XML hasorganizational Processes: been widely accepted as a protocol for ex-Aalst and Kumar changing business data, there is still no clear standard for exchanging business process infor-Workflow models are an effective means for de- mation (e.g., workflow models). This is the veryscribing, analyzing, implementing, and managing relevant problem addressed by this research. MIS Quarterly Vol. 28 No. 1/March 2004 93
  20. 20. Hevner et al./Design Science in IS ResearchResearch Rigor Design as an ArtifactResearch on workflow modeling has long been There are two clearly identifiable artifacts pro-based on rigorous mathematical techniques such duced in this research. First, the workflow lan-as Markov chains, queueing networks, and Petri guage XRL is designed. XRL is based on Petri-nets (Aalst and Hee 2002). In this paper, Petri net formalisms and described in XML syntax.nets provide the underlying semantics for XRL. Interorganizational business processes areThese formal semantics allow for powerful analy- specified via XRL for execution in a distributed,sis techniques (e.g., correctness, performance) to heterogeneous environment.be applied to the designed workflow models.Such formalisms also enable the development of The second research artifact is the XRL/flowerautomated tools to manipulate and analyze com- workflow management architecture in which XRL-plex workflow designs. Each language construct described processes are executed. The XRLin XRL has an equivalent Petri-net representation routing scheme is parsed by an XML parser andpresented in the paper. The language is exten- stored as an XML data structure. This structure issible in that adding a new construct simply read into a Petri-net engine which determines therequires defining its Petri-net representation and next step of the business process and informs theadding its syntax to the XRL. Thus, this research next task provider via an e-mail message. Resultsdraws from a clearly defined and tested base of of each task are sent back to the engine whichmodeling literature and knowledge. then executes the next step in the process until completion. The paper presents a prototype implementation of the XRL/flower architecture asDesign as a Search Process a proof of concept (Aalst and Kumar 2003).XRL is designed in the paper by performing a Another artifact of this research is a workflowthorough analysis of business process require- verification tool named Wolfan that verifies thements and identifying features provided by leading soundness of business process workflows.commercial workflow management systems. Soundness of a workflow requires that theUsing the terminology from the paper, workflows workflow terminates, no Petri-net tokens are lefttraverse routes through available tasks (i.e., behind upon termination, and there are no deadbusiness services) in the electronic business tasks in the workflow. This verification tool isenvironment. The basic routing constructs of XRL described more completely in a different paperdefine the specific control flow of the business (Aalst 1999).process. The authors build 13 basic constructsinto XRL: Task, Sequence, Any_sequence,Choice, Condition, Parallel_sync, Parallel_no_ Design Evaluationsync, Parallel_part_sync, Wait_all, Wait_any,While_do, Stop, and Terminate. They show the The authors evaluate the XRL and XRL/flowerPetri-net representation of each construct. Thus, designs in several important ways:the fundamental control flow structures ofsequence, decision, iteration, and concurrency are • XRL is compared and contrasted with lan-supported in XRL. guages in existing commercial workflow systems and research prototypes. TheThe authors demonstrate the capabilities of XRL majority of these languages are proprietaryin several examples. However, they are careful and difficult to adapt to ad hoc businessnot to claim that XRL is complete in the formal process design.sense that all possible business processes can bemodeled in XRL. The search for a complete set of • The fit of XRL with proposed standards isXRL constructs is left for future research. studied. In particular, the Interoperability Wf-94 MIS Quarterly Vol. 28 No. 1/March 2004
  21. 21. Hevner et al./Design Science in IS Research XML Binding standard (WfMC 2000) does not Information Systems Design for at this time include the specification of control Emergent Knowledge Processes: flow and, thus, is not suitable for inter- Markus, Majchrzak, and Gasser organizational workflows. Electronic com- merce standards (e.g., RosettaNet) provide Despite decades of research and development some level of control flow specification for efforts, effective methods for developing infor- predefined business activities, but do not mation systems that meet the information require- readily allow the ad hoc specification of ments of upper management remain elusive. business processes. Early approaches used a “waterfall” approach where requirements were defined and validated• A research prototype of XRL/flower has been prior to initiating design efforts which, in turn, were implemented and several of the user interface completed prior to implementation (Royce 1998). screens are presented. The screens demon- Prototyping approaches emerged next, followed strate a mail-order routing schema case by numerous proposals including CASE tool- study. based approaches, rapid application development, and extreme programming (Kruchten 2000).• The Petri-net foundation of XRL allows the Walls et al. (1992) propose a framework for a authors to claim the XRL workflows can be prescriptive information system design theory verified for correctness and performance. aimed at enabling designers to construct “more XRL is extensible since new constructs can effective information systems” (p. 36). They apply be added to the language based on their this framework to the design of vigilant executive translation to underlying Petri-net repre- information systems. The framework establishes sentations. However, as discussed above, a class of user requirements (model of design the authors do not make a formal claim for problems) that are most effectively addressed the representational completeness of XRL. using a particular type of system solution (instantiation) designed using a prescribed set of development practices (methods). Markus et al.Research Contributions (2002) extend this framework to the development of information systems to support emergentThe clear contributions of this research are the knowledge processes (EKPs)—processes indesign artifacts—XRL (a workflow language), which structure is “neither possible nor desirable”XRL/flower (a workflow architecture and its (p. 182) and where processes are characterizedimplemented prototype system), and Wolfan (a by “highly unpredictable user types and workPetri-net verification engine). Another interesting contexts” (p. 183).contribution is the extension of XML in its ability todescribe and transmit routing schemas (e.g.,control flow information) to support interorgani- Problem Relevancezational electronic commerce. The relevance and importance of the problem are well demonstrated. Markus et al. describe a classResearch Communication of management activities that they term emergent knowledge processes (EKPs). These includeThis paper provides clear information to both “basic research, new product development,technical and managerial audiences. The presen- strategic business planning, and organizationtation, while primarily technical with XML coding design” (p. 179). They are characterized by “pro-and Petri-net diagrams throughout, motivates a cess emergence, unpredictable user types andmanagerial audience with a strong introduction on use contexts, and distributed expert knowledge”risks and benefits of applying interorganizational (p. 186). They are crucial to many manufacturingworkflows to electronic commerce applications. organizations, particularly those in high-tech MIS Quarterly Vol. 28 No. 1/March 2004 95