Assessing the complexity of BPMN models for redesign using cluster analysis

Abstract

Business Processes exist at the core of each organisation and their efficient management is a main objective for those aiming to benefit from a process-centric approach. Today’s rapidly changing economic environment introduces the challenge of analysing, maintaining and optimizing increasingly complex processes. Increased complexity is considered to have a negative impact on the success of Business Process Redesign initiatives. Several aspects of Business Process complexity have been studied in literature, mostly focussing on complexity measurement and proposal of appropriate complexity metrics. The present research leverages metrics introduced in research for the development of complexity assessment methods. These methods aim at providing a straightforward way of evaluating a process model’s complexity to draw conclusions regarding a model’s capability for Redesign. Through the implementation of cluster analysis on BPMN models, based on selected complexity metrics, the identification of highly complex models becomes feasible. The developed methods utilize the K-means algorithm for clustering in order to create a model for distinguishing between complexity levels. The first proposed approach provides reference values for the categorization of future Business Process models, while a second approach combines complexity metrics to a single weighted sum measure. This approach is presented with the purpose of simplifying the assessment of models by the definition of exact thresholds, as well as offering a means of assigning priorities to complexity metrics. The latter is found to aid in the identification of problematic areas on a process model, by examining the impact of each metric on the overall complexity.