What is process mining? Explained

What process mining does

Every time someone updates a ticket, approves an invoice or closes an order, your software leaves a timestamped record. Process mining reads those records and rebuilds the path each case took.

IBM describes it as applying specialized algorithms to event log data to find trends, patterns and details of how a process unfolds. It is often called business process mining when applied to company workflows.

Wil van der Aalst, a Dutch computer scientist, is credited with much of the academic research behind the field.

Process mining vs process mapping

A process map shows how work is supposed to flow.

FEMA describes it as a visual of an existing single process from start to finish, usually drawn from interviews and workshops. Read more in process mapping.

Process mining starts from the data instead. It shows how cases actually moved, including the shortcuts and loops nobody drew.

The two work well together: map the intended flow, then mine the logs to see where reality differs. Our process mapping examples show the first half.

What it finds and where it falls short

Once the real flow is visible, patterns stand out. Teams use it to spot slow steps, repeated work and the many different routes a single process takes.

IBM also notes a limit: manual tasks outside IT systems are not recorded in event logs, so process mining can miss them. Work done in a spreadsheet, on a call or on paper stays invisible.

Treat the result as a strong view of system activity, not the whole picture.

• Delays where cases wait the longest
• Rework where steps repeat
• Variants, meaning the different paths cases take
• Blind spots in manual work outside the systems

How process mining feeds automation

• Find the waste

  Spot delays and rework in real data.

• Fix the process

  Redesign before you automate.

• Pick candidates

  Choose stable, repeated steps.

• Automate

  Build the workflow on what works.

Automating a broken process only makes it fail faster. Michael Hammer argued in Harvard Business Review that firms should redesign processes rather than embed old ones in software.

Process mining helps you pick the right candidates. A step that is stable, repeated often and low on exceptions is a good fit, a point echoed in research on robotic process automation.

From there, you can move to workflow automation or business process automation, with the data showing what to fix first.