Process mining is a method for analyzing how a process operates by using system event logs to reconstruct the sequence of activities for individual cases, such as purchase orders or insurance claims.
An event log requires a case identifier, an activity name, and a timestamp for each recorded event. In a DMAIC project, the case definition and measurement boundary determine which records belong in the analysis. Teams use the resulting process paths to examine departures from the expected workflow and locate repeated steps that warrant investigation.
Consider an invoice process in which 120 of 1,000 completed cases contain a second approval event. Process mining identifies those cases and shows where the repetition occurs. If review confirms that only 80 cases required correction, the case-level rework proportion is 8%. The remaining repetitions might reflect authorized approval requirements. Counting every repeated event as a defect would overstate the problem.
Timing analysis requires the same care. A two-day gap between invoice entry and approval may include a weekend or a scheduled batch release. Completion timestamps alone do not establish how much time an invoice spent waiting for an available approver. A capacity decision needs evidence about resource availability and release rules.
Within DMAIC, process mining supports Measure and Analyze, while SIPOC establishes scope and a gemba walk exposes work absent from system records. Ishikawa diagrams and 5 Whys help investigate explanations suggested by the logs. Control plans should assign ownership of event definitions because a changed timestamp or activity mapping can break comparability with the baseline.
