Rigor and Relevance—A Design-Science Story about BPMN

For many years, I have had an on-again, off-again relationship with design science. I read the first edition of Herbert Simon’s book The Sciences of the Artificial shortly after it was published, and I was greatly attracted to the notion of a science of design that Simon proposed. Indeed, for a number of years, I used his book as required reading in an information systems course I taught. I have always felt that viewing information systems through the lens of designing an artifact lies at the foundation of the information systems field. Simon’s book was one of the few that provided a scholarly foundation to support such a perspective. I continue to struggle, however, to come to a deep understanding of what constitutes high-quality design-science research.

In this regard, a paper that was not framed as a piece of design-science research captures for me the essence of high-quality design-science research. I first learned about the paper because it used a theory proposed by Yair Wand and me in the early 1990’s as a basis for evaluating conceptual-modeling grammars. The paper by Recker et al.1 uses this theory to predict that practitioners would face nine difficulties when they used the Business Process Modeling Notation (BPMN) for process modeling. They then undertook a field study in which they interviewed 19 practitioners to determine whether their theoretical predictions had empirical support.

Importantly, as an exemplar of research that is rigorous and relevant, Recker et al. subsequently discussed their findings with the lead designer of BPMN. He acknowledged some of the difficulties they reported. Moreover, he indicated that future versions of BPMN might address them. Interestingly, later versions of BPMN removed some constructs that Recker et al.’s research indicated were problematic (e.g., the Off-Page Connector construct) and included other constructs that earlier versions of BPMN lacked but were deemed necessary in the research findings (e.g., a Business-Rule construct). Contrary to the research findings, however, later versions of BPMN increased the number of constructs that could be used to represent events, which anecdotally appears to have exacerbated modelers difficulties in using BPMN (consistent with Recker et al.’s research findings).

Why do I believe Recker et al.’s paper constitutes an example of high-quality design-science research? First, it uses a “kernel theory” to make predictions about the strengths and weaknesses of an important information technology artifact—one used extensively in practice. Second, it provides design guidelines that developers can and prima facie have employed to try to improve the artifact. As a prototypical piece of design-science research, perhaps what is lacking from Recker et al.’s work is evidence of design iteration. This requirement could be addressed, however, if a research program were to replicate their work with later versions of BPMN and the findings again provided to the developers of BPMN to inform their subsequent design work.

1. Jan Recker, Marta Indulska, and Peter Green, “Extending Representational Analysis: BPMN User and Developer Perspectives,” in G. Alonso, P. Dadam, and M. Rosemann (Eds.), BPM 2007, LNCS 4714, Springer-Verlag Berlin Heidelberg 2007, pp. 384–399, 2007.