Process Mining: Data-driven process discovery

Process Mining: Data-Driven Discovery

Wil van der Aalst 2000s High Complexity

Process Mining is a data-driven technique that extracts process knowledge from event logs in information systems, revealing how processes actually work versus how they're documented.

What Is It?

Process Mining bridges the gap between traditional process analysis (based on interviews and documentation) and data analytics. Instead of asking how processes should work, it discovers how they actually work by analyzing event logs from systems like ERP, CRM, and workflow tools.

There are three types of Process Mining: Discovery creates a process model purely from event log data—no prior model needed. Conformance compares actual behavior to a reference model to find deviations. Enhancement enriches models with performance data like bottlenecks, delays, and resource utilization.

The insights are often surprising. Organizations frequently discover that actual processes differ significantly from documented procedures—with hidden rework loops, undocumented workarounds, and unexpected bottlenecks. This evidence-based approach removes opinion and politics from process improvement discussions.

Process Mining is often used within Business Process Management initiatives and pairs well with Value Stream Mapping for visualization.

Process Mining: From event logs to discovered process
Process Mining: Extracting process patterns from event logs

Quick Reference

Complexity
High (8/10)
Time to Decision
2-4 weeks
Data Required
Very High
Team Size
2-10
Objectivity
Very High
Learning Curve
4-8 weeks

Core Features

  • Discovery: Create process models automatically from event logs
  • Conformance: Compare actual vs. documented processes
  • Enhancement: Add performance data to process models
  • Bottleneck Analysis: Identify where delays occur
  • Variant Analysis: See different paths through processes
  • Root Cause Analysis: Find why deviations occur

When to Use

  • Need objective, evidence-based process understanding
  • Processes are complex with many variants
  • Systems have quality event log data available
  • Current documentation doesn't match reality
  • Compliance or audit requirements
  • Within BPM initiatives for discovery

When NOT to Use

  • No event log data available or poor data quality
  • Simple processes that don't need data analysis
  • Rapid improvement needed (use Kaizen Blitz)
  • Manual processes without system tracking
  • Organizations without analytical capabilities

Key Strengths

  • Evidence-Based: Removes opinion and politics from analysis
  • Scalable: Analyzes thousands of process instances
  • Objective: Data-driven insights, not assumptions
  • Revealing: Discovers hidden inefficiencies and deviations
  • Continuous: Can monitor processes in real-time

Key Weaknesses

  • Requires quality event log data (garbage in, garbage out)
  • Specialized tools and skills needed
  • Data preparation can be time-consuming
  • Results need interpretation by process experts
  • May reveal uncomfortable truths about processes

How It Works

1 Primary InputEvent logs with Case ID, Activity, Timestamp (and optionally resources, attributes)
2 Data You NeedSystem event logs, reference process models (for conformance), performance targets
3 Primary OutputDiscovered process models, conformance reports, bottleneck analysis, improvement opportunities

Comparison with Related Frameworks

Process Mining vs Business Process Management

BPM is the full lifecycle discipline. Process Mining is a specific analytical technique. Process Mining is often used within BPM to discover and monitor processes.

Process Mining vs Value Stream Mapping

Value Stream Mapping creates manual visualizations through observation. Process Mining discovers processes from data. VSM is qualitative; Process Mining is quantitative. They complement each other.

Deep Resources