What is process mining?

Process mining is a data science technique that extracts a business process's real execution path from the event logs already produced by enterprise systems — ERPs, case-management tools, ticketing systems — using timestamps and case IDs to algorithmically reconstruct the sequence, timing, and variants of the process as it actually happened.

Process mining starts from an event log: a table of records where each row has a case identifier, an activity name, and a timestamp, usually exported from a system such as an ERP, a ticketing tool, or a case-management platform. A process-mining algorithm groups the events by case ID, orders them by timestamp, and looks for the sequences that repeat across cases. From that it draws a process map showing the paths cases actually took, the frequency of each path, and how long each step and transition took. Because the technique is purely a function of the log's structure and completeness, its output is only as good as the data the underlying systems were set up to record.

The technique depends on the source systems already emitting structured, timestamped, case-identified events — which is common in ERP and ticketing software but rare in the ad hoc mix of spreadsheets, email threads, chat messages, and manual steps that make up much of real office work. Where that structured logging does not exist, process mining has nothing to mine, and the process stays invisible even though the work is still happening.

Process mining reconstructs a process from the event logs systems already keep. Task mining instead observes the user's screen — clicks, keystrokes, application switches — to capture the steps a person takes, including the parts of the work that never touch a loggable system. The two are complementary: process mining is strong wherever clean system logs already exist, and task mining is strong wherever the real work is manual, cross-application, or otherwise undocumented by any single system's logs.