Data Quality Assessment to Apply Process Mining in Production Processes

The modern industrial environment is characterized by short product lifecycles and high competitive pressure. Therefore, efficient production processes are highly important to stay competitive. Due to this, the interest in Business Process Management (BPM) reached more importance over the last years. The first step in BPM is process discovery. However, traditional workshop-, interview- or evidence-based methods (e.g. value stream mapping) are subjective and time-consuming. To overcome these disadvantages, Process Mining (PM) can be used to describe processes unbiased and fact-based. PM uses process feed-back data from existing enterprise software, which are stored in event logs, to describe production processes. The quality of the resulting process model strongly depends on the data quality from the systems. Because of the broad variety of historically grown IT solutions in production industries, the quality of data is often too heterogeneous to be used. There are also frequent problems in reported data, such as missing order identification or different time stamps for the same event. To address these problems, PM-specific data quality problems are identified in this paper and an approach to assess the quality of event log data in production is presented. Additionally, further research outlooks for a subsequent data enhancement are given, in order to apply PM in production successfully.