Empirical Studies in Process Model Verification

Despite the large body of knowledge on formal analysis techniques for process models, in particular Petri nets, there has been a notable gap of empirical research into verification. In this paper we compare the few studies that report results from applying verification techniques to real-world process model collections. For this comparison we are particularly interested in the different approaches, their computational performance, and the number of errors found. Our comparison reveals that most of the samples have error rates of 10% to 20%. Some of the studies have established a connection between error probability and process model metrics, as well as between model understanding and both metrics and modeling competence of the model reader. Based on these results, we discuss implications and directions for future research.