What is Process Mining?
Process mining is an application of data mining with the aim of mapping and discovering an organization’s processes. While data mining leverages big data to gain insights on business performance, process mining provides an understanding of the overarching processes within an organization.
Process mining is used to better understand the real (non-hypothetical) processes that actually exist within systems. When combined with a simulation modeling tool like HASH, this has applications in:
- improving the efficiency of operations;
- identifying weaknesses within processes;
- identifying undocumented processes and procedures that may represent a risk in the event of unexpected worker absence or leave.
Through process mining, areas of risk and opportunity can be systematically identified, and reviewed.
How does Process Mining work?
Process mining was historically something of a manual task which involved analysts (often outside consultants) shadowing and observing workers and departments to build a picture of their everyday day-to-day. This approach was expensive and time-consuming, and came with the risk of an observer effect altering the validity of results.
Modern process mining utilizes discrete software and analytics programs to map worker activities and their relation to one another, in an automated, silent fashion. These software packages keep ‘event logs’ each and every time a process is completed, along with relevant information as to how that process was carried out (e.g. what applications or webpages were open? which pieces of data were pasted from one application to another? how long did it take to complete in total?)
From these event logs, performance metrics can be determined, and underlying real-world practices better understood. Discrepancies between reality on the ground, and formally documented processes can be easily detected (allowing for investigation as to why), and non-performant or redundant processes addressed. Through continuous monitoring and process mining, KPIs and targets for improvement can be set and effectively pursued.
Types of Process Mining
There are three main types of process mining, which are dependent upon whether an existing process model can be used.
Conformance checking or performance analysis are used when existing models exist, whereas discovery is used in the absence of a prior model.
- Conformance checking: event logs are compared to the prior model and are analyzed for points of adherence and divergence. Performance data can reveals any aberrations from a reference model, and may be used to forecast future potential bottlenecks under varying conditions, through Monte Carlo-style simulation experiments in tools such as hCore. Decision-makers can use these insights to adjust or overhaul processes to ensure future conformance to the model.
- Performance analysis: used not to check existing data against the model, but to optimize the performance of the reference model itself, based on a number of performance objectives (e.g. level of standardization, relative cost, cycle times and quality of end product — or stakeholder satisfaction).
- Discovery: a didactic approach to process mining, where a new model is constructed based on event logs. This is especially useful where processes exist but may not be known or well-understood.
How is Process Mining used?
Process mining can unveil hidden causal relationships in a system, where underperformance in one area harms outcomes in another. These interdependencies can then be captured and explored in more detail in a simulation environment such as HASH. This is especially effective in modeling dynamic and variable real operational processes, where digitized (yet unstructured) process data exists.
Below are the main uses of process mining in business process modeling:
- Process monitoring for security and adherence: the act of logging/recording information about processes as they occur for investigation, warning and audit purposes.
- Process analysis for resiliency: using log files to identify strengths, flaws, redundancies and deficiencies in process operations.
- Process design optimization: using insights gained from process monitoring to adjust or reinvent existing processes.
- Process identification: data mining algorithms are applied to event logs to identify patterns, correlations and anomalies in event data and document processes not already known to exist.
Troubleshooting and then re-engineering processes helps with operational management, also allowing the user to monitor the impact of process improvements after they have been made. Meanwhile, fully understanding existing processes and increasing transparency of workflows present opportunities for strategic shifts, or to pivot based on recent process developments. It allows organizations to maintain a competitive edge, through taking a long-term view on their internal processes.