Quality of process modeling using BPMN: a model-driven approach

Abstract

Context: The BPMN 2.0 specification contains the rules regarding the correct usage of the language’s constructs. Practitioners have also proposed best-practices for producing better BPMN models. However, those rules are expressed in natural language, yielding sometimes ambiguous interpretation, and therefore, flaws in produced BPMN models. Objective: Ensuring the correctness of BPMN models is critical for the automation of processes. Hence, errors in the BPMN models specification should be detected and corrected at design time, since faults detected at latter stages of processes’ development can be more costly and hard to correct. So, we need to assess the quality of BPMN models in a rigorous and systematic way. Method: We follow a model-driven approach for formalization and empirical validation of BPMN well-formedness rules and BPMN measures for enhancing the quality of BPMN models. Results: The rule mining of BPMN specification, as well as recently published BPMN works, allowed the gathering of more than a hundred of BPMN well-formedness and best-practices rules. Furthermore, we derived a set of BPMN measures aiming to provide information to process modelers regarding the correctness of BPMN models. Both BPMN rules, as well as BPMN measures were empirically validated through samples of BPMN models. Limitations: This work does not cover control-flow formal properties in BPMN models, since they were extensively discussed in other process modeling research works. Conclusion: We intend to contribute for improving BPMN modeling tools, through the formalization of well-formedness rules and BPMN measures to be incorporated in those tools, in order to enhance the quality of process modeling outcomes.