Detecting the “Split-Cases” Workaround in Event Logs

Workarounds are frequently observed in business processes, where employees intentionally bypass or deviate from procedures and business rules embedded in process models. One of the common workaround behaviors is the “split cases” workaround. This workaround typically takes place in processes where threshold conditions over data items are used by decision rules, determining the process path to be followed. Being familiar with the process rules, employees may decide to split a case whose relevant data value exceeds the threshold into two or more cases whose data value is below the threshold. Doing this they avoid the path which should be followed otherwise – this path may seem undesirable to the employee (e.g., including lengthy approvals), but might be needed for avoiding risks. Detecting such workarounds is hence of importance. Although the split case workaround is quite common, existing process mining and conformance checking techniques are not geared to detect it in an event log, since each of the cases is conformant by itself, and it is the relation among cases which might indicate this behavior. In this paper we propose an approach for detecting the split cases workaround in event log and report experimentation that has been performed using simulated as well as real life logs.