Business processes executed in organisations are increasingly recorded by information systems. These collections of data (implicitly) describe events (e.g., placing an order or taking a blood test) and, hence, provide information on the actual execution of business processes. This log data often not only contains information about process execution, but also about the context of the executed process, allowing the event log to be split into different process cohorts based on these attributes (e.g. “all patients that arrived before 6pm”). Analysing differences between different process cohorts can then reveal positive and/or negative effects of context factors on process execution and deliver valuable insights into how management of a process can be improved. The amount of data however is usually so big, that it is difficult to identify such differences across many attributes and different steps of a process.
The goal of this research is to develop a new data visualisation framework that allows stakeholders to quickly and efficiently identify differences between process cohorts in their processes.
The key questions in this research include:
- How can we visualize data captured in an event log in such a way that enables stakeholders to extract relevant and accurate information about differences between process cohorts in a fast and efficient manner?
- How can we provide both high-level overviews of the data and low-level detail to enable an intuitive detection of anomalies, trends and root-causes?
In collaboration with researchers from Eindhoven University of Technology, we have managed to develop and implement a visualisation framework that enables the visualisation of performance indicators for multiple process cohorts across the entire business process. This enables process stakeholders to visually detect differences between process cohorts quickly and intuitively.
In particular, our framework allows the computation of performance statistics for process cohorts at all steps of a process and the visualisation of all these statistics at once, providing an overview over the all states across the entire process at one glance. This framework is customisable in how a process cohort’s performance is visualised: one can select one or several projections of statistics on to a process map that provides both a high-level overview and low-level detail of the performance statistics of multiple process cohorts for the executed process.
This framework has been implemented as a plug-in to the open-source ProM tool (called “ProcessProfiler3D”).
Wynn, M. T., Poppe, E., Xu, J., ter Hofstede, A. H., Brown, R., Pini, A., & van der Aalst, W. M. P. (2017). ProcessProfiler3D: A Visualisation Framework for Log-based Process Performance Comparison. Decision Support Systems.