I have walked into certified plants where the documented ISO 9001 procedure claims a customer complaint reaches root cause analysis in 48 hours. When you actually time-stamp the handovers, the average closure takes 14 days. The documentation is not a deliberate lie; it is a structural illusion built on what people believe they do.
People do not follow the procedure. They follow what works. What works usually involves system workarounds, waiting for informal approvals, and routing steps that nobody documents. You will never find these hidden loops in a SIPOC diagram or a standard workshop.
Process mining is the mechanism that closes the gap. By extracting the digital footprints left in enterprise IT systems, the technology reconstructs the actual process flow. It replaces subjective opinions with timestamped reality.
The Limits of Traditional Process Mapping
The traditional quality engineering toolkit relies on methods designed to capture intent. We build flowcharts to define how things should operate. We construct Value Stream Maps based on timed observations during a scheduled Gemba walk.
Turtle diagrams map inputs and outputs through interviews with process owners. SIPOC diagrams define scope during management workshops. Every one of these tools is valuable for establishing a baseline, but they share a critical flaw: they are entirely subjective.
These maps reflect what people think they do, not what the system records. When an auditor asks for evidence of conformance, showing a hand-drawn flowchart is insufficient. Auditors compare documentation against operational reality, and traditional mapping cannot quantify that gap.
Process mining solves this subjectivity. It does not ask operators how they execute a task. It looks directly at the ERP, MES, and QMS logs to show exactly what happened, when it happened, and who executed it.
Process Mapping vs. Process Mining
Traditional Mapping (Intent)
- Relies on interviews and manual observation
- Shows the happy path and ideal flow
- Fails to capture informal workarounds
- Provides a static snapshot in time
Process Mining (Reality)
- Extracts 100% of system timestamp data
- Highlights deviations and hidden loops
- Quantifies waiting times between steps
- Continuously monitors process conformance
Anatomy of an Event Log
Every process mining initiative begins with the event log. This is a structured data table where each row represents a single system event. Modern platforms extract this data directly from QMS and ERP databases without manual interference.
To build a valid model, you need four specific data points for every transaction. Without a reliable Case ID, the algorithm cannot group events into individual instances. Without precise Timestamps, the tool cannot calculate throughput or waiting times.
The Activity column records the specific action taken, such as opening an 8D or approving a deviation. The Resource column identifies the user or system node that executed it. If these four fields are clean, the algorithm can reconstruct the entire operational landscape.
Data extraction frequently exposes immediate system flaws. During my projects, we often find missing timestamps or duplicate Case IDs during the initial extraction. Cleaning the log to enable mining is itself a diagnostic exercise that highlights systemic data governance issues.
| Data Field | Function | Example Entry |
|---|---|---|
| Case ID | Links events to a specific instance | REC-2026-0421 |
| Activity | Identifies the executed step | Root Cause Analysis |
| Timestamp | Provides exact chronological order | 2026-04-21 14:32:07 |
| Resource | Identifies the person or system | J. Smith / QMS-Bot |
Deconstructing a Blocked Complaint Process
I deployed process mining at an automotive supplier facing a crisis in customer complaint handling. Their documented procedure listed 12 clean steps. The corporate KPI showed an average 8D closure time of 23 days against a customer-mandated target of 10 days.
The OEM was threatening PPAP revocation. A traditional approach would have meant locking the quality team in a room to map the bottlenecks. Instead, we extracted 12 months of data from their QMS, capturing 47,000 individual events across 3,200 distinct complaints.
We loaded the data into Fluxicon Disco for exploratory analysis. The algorithm immediately generated a reality map that looked nothing like the ISO 9001 documentation. The process did not have 12 steps; it had 38 distinct variants ranging from 15 to 27 steps.
The most frequent path, covering 65 percent of cases, contained a massive hidden bottleneck. Complaints were sitting in an undocumented queue waiting for controlling to approve a financial reservation. That single unauthorised step averaged 6.3 days per file.
Conformance Checking and the 8D Bottleneck
The mining tool also exposed severe routing failures. In 12 percent of cases, the QMS automatically assigned the complaint to the wrong engineering team. The file would sit untouched for four days before someone manually redirected it back to the correct department.
Furthermore, 8 percent of 8D reports underwent redundant reviews. The system lacked clear ownership rules, resulting in one review by an internal auditor and a second approval by the quality manager. This duplicated effort wasted resources and delayed closure.
This is why conformance checking is the most powerful feature for quality engineers. It overlays the actual process data directly onto the target ISO or IATF standard. It quantifies the exact deviation rate, showing you skipped steps and unauthorised additions.
Process mining gives you the evidence to fix the system, not just the means to discipline an operator.
We eliminated the controlling approval for complaints under 500 EUR, which covered 80 percent of the volume. We implemented automated routing based on the failure code and standardised the 8D review to a single level. The average closure time dropped from 23 days to 9 days in less than three months.
Integrating Mining with Core Quality Metrics
Process mining is not limited to administrative 8D workflows. It directly impacts manufacturing metrics like OEE and Process Capability. If your Cpk calculations look stable during the day shift but degrade at night, process mining can correlate those drops to actual workflow changes.
When you map the MES data, you often find that night-shift operators are forced to deviate from the standard work instructions due to material shortages or missing maintenance approvals. The variability in your output is not just a machine issue; it is a process flow issue.
Supplier quality management also benefits heavily. By mining the ERP data, you can track the exact time between raising a supplier claim and their formal 8D response. It replaces anecdotal complaints with hard data regarding supplier lead times and material handling delays.
Finally, it transforms audit preparation. Before an EASA or IATF 16949 audit, conformance checking proves to the auditor that you manage compliance systematically. You are not showing them a random sample; you are showing them 100 percent population data.
Deploying a Process Mining Pilot
- 01Select a Target ProcessChoose a high-volume, digital workflow like customer complaints with a clear owner.
- 02Extract and Validate DataPull event logs from the QMS and verify Case IDs, timestamps, and resources.
- 03Run Discovery AnalysisGenerate the actual map to identify loops, deviations, and primary bottlenecks.
- 04Execute Conformance CheckingCompare actual steps against documented procedures and ISO standards.
- 05Implement and MonitorDeploy process fixes and use dashboards to verify the cycle time reduction.
Avoiding Deployment Failures
The most common mistake is treating process mining as an IT deployment. IT can extract the database tables and configure the software, but they cannot interpret the quality data. If you leave process mining to IT, you will generate beautiful dashboards with zero operational impact.
Do not wait for perfect data to start. Nobody has perfectly clean event logs. Even incomplete data will reveal your most significant systemic bottlenecks. The act of cleaning the data often exposes underlying governance issues that the quality team needs to address anyway.
Process mining does not replace Lean or Six Sigma. Lean provides the philosophy of waste elimination, and Six Sigma provides the statistical rigour for variation reduction. Process mining provides the objective baseline that makes both methodologies significantly more effective.
As digitalisation accelerates under Industry 4.0, the volume of event data will only increase. Integrating artificial intelligence with process mining enables predictive monitoring. The system will flag a complaint that has a 78 percent probability of breaching a 15-day limit, allowing you to intervene before the failure occurs.