Quality management is not a personality test. Over twenty-plus years across automotive and aerospace plants, I have seen the same failure modes repeat: systems designed for the auditor, not the operator; KPIs that measure activity instead of outcome; and corrective actions that close the 8D without fixing the process. The plants that break this cycle share one trait — they treat the standard as a floor, not a ceiling.
The work itself is unglamorous. It means standing in a stamping shop at six in the morning, watching a press cycle, and reconciling what the PFMEA says should happen with what actually happens. It means building a greenfield QA department for over 900 people who have never worked under IATF 16949, and convincing production that containment is not punishment. The frameworks — ISO 9001, AS9100, VDA 6.3 — are the easy part. Execution is where it fails.
What follows is a set of practical lessons from building, transitioning, and rescuing quality systems. No motivational theory. Just the mechanisms, the numbers, and the failure modes I have encountered implementing this work at a major aerospace manufacturer, SNOP, WITTE Automotive, and earlier at Honeywell and Molex.
The Framework Problem: Compliance Versus Capability
Most quality systems are built backwards. A team takes the ISO 9001 or IATF 16949 standard, maps its clauses to a set of procedures, and declares the system operational. The audit passes. Six months later, nothing has changed on the shop floor. I have audited plants that held pristine documentation while their first-pass yield bled red. The standard gives you the framework for a house; it does not build the house.
The fix is to design the quality system around process capability, not clause compliance. Start with the process flow. Identify where variation enters. Build your control plan around those specific failure modes. Then — and only then — check whether your documentation satisfies the auditor. When you reverse the order, you get a system that looks correct on paper and collapses under real production load. I have transitioned ISO 9001 systems at three major manufacturers, and the pattern holds every time.
The practical test is simple. Walk the gemba with the production supervisor. Ask them to show you the last three deviations on their line. If the operator cannot explain the reaction plan, your control plan is theatre. If the supervisor blames the quality department for the deviation, ownership has not been established. These are system failures, not people failures.
Two Approaches to System Design
Compliance-first (typical)
- Map standard clauses to written procedures
- Audit documentation for conformity
- Build control plan from PFMEA template
- Quality owns deviation closure on paper
Capability-first (effective)
- Walk the process and map variation entry points
- Design reaction plans operators can execute
- Build PFMEA from actual failure data
- Production owns the process; quality verifies
Routing Verification: Cutting Lead Time by Changing the Question
At a major aerospace manufacturer, internal lead time for routing verification was choking the value stream. The standard approach was to inspect at the end, log nonconformities, and route them back through a rework loop. The system generated enormous queues. The KPI measured how fast the quality team cleared the backlog. The problem was that the KPI measured the wrong thing.
I introduced a set of Routing Verification KPIs that shifted the measurement from backlog clearance to first-time-right routing. Instead of tracking rework velocity, we tracked the percentage of routings that passed verification without a single deviation. The change was mechanical, not philosophical. We moved the inspection point upstream, gave immediate feedback to the source, and made the deviation visible before it entered the queue.
The result was a 97% reduction in internal lead time. That number sounds implausible until you decompose it. The old system spent most of its time managing, prioritising, and re-entering parts into the flow. By eliminating the rework loop at its origin, we removed the queue itself. You cannot optimise your way out of a bad process architecture. You have to change the architecture.
Building a Greenfield QA Department: Structure Before Staff
Building a quality department from scratch for a plant with over 900 employees — which I did at SNOP — teaches you that structure precedes staffing. The temptation is to hire experienced inspectors and let them figure out the system. That approach guarantees inconsistency. Every inspector brings their own interpretation of acceptance criteria, their own documentation habits, and their own threshold for when to escalate.
The first step is to define the governance layer: who approves deviations, who signs off on PPAP submissions, who owns the supplier scorecard. Then define the operational layer: incoming inspection, in-process controls, final audit. Only then do you staff it. I structured the department around the process flow, not around the org chart. Each quality function mapped to a specific production stage with a clear handoff.
The critical success factor was Measurement Systems Analysis (MSA) before the first production run. If your gauges are not capable, your inspection data is noise. We ran gauge R&R studies on every critical-to-quality characteristic, established calibration intervals to ISO 17025 requirements, and locked the measurement plan before the launch team began building parts. This eliminated the most common launch failure: rejecting good parts because the measurement system itself was unstable.
Supplier Quality: The Leverage You Are Not Using
In my consulting years, the most common request was help with supplier nonconformities. Plants were spending fortunes on incoming inspection, sorting, and supplier chargebacks. The root cause was almost never a bad supplier. It was a bad supplier management process. Teams were auditing suppliers against generic checklists, approving PPAP packages that were incomplete, and then acting surprised when the production parts did not match the samples.
Effective supplier quality starts with the APQP process. If you do not align your requirements with the supplier's process FMEA before tooling is cut, you will spend the entire production life arguing about deviations. The VDA 6.3 process audit is the right tool, but it must be applied at the supplier's site during the ramp-up phase, not after the first failure. I have seen plants reduce supplier PPM by an order of magnitude simply by moving the audit upstream and tying the results to the release of production volume.
If your supplier audit is a post-mortem, you have already paid for the failure.
The escalation path matters as much as the audit. Define exactly what triggers a supplier corrective action request, what evidence is required in the 8D response, and at what point you invoke capacity hold. Too many teams escalate emotionally — shouting when a delivery is late, but ignoring systemic drift on a critical dimension. A calm, documented escalation matrix removes the emotion and forces the data to drive the conversation.
Supplier Escalation Sequence
- 01Level 1: Deviation loggedSCAR issued with full containment and 8D D1-D3 within 48 hours
- 02Level 2: Systemic recurrenceOn-site VDA 6.3 process audit; results tied to volume release
- 03Level 3: Capability failureCapacity hold imposed; new product sourcing initiated
- 04Level 4: ExitResourcing plan executed; PPAP for replacement supplier fast-tracked
Capability Indices: What the Number Actually Tells You
A Cpk of 1.33 is the minimum acceptable threshold in most automotive and aerospace environments. But the number alone is meaningless without context. I have reviewed capability studies where the index was calculated across multiple cavities, multiple machines, and multiple shifts — producing a composite number that masked individual process failures. The composite looked fine. Cavity three was producing scrap.
Capability must be calculated per independent process stream. If you have a four-cavity mould, you need four capability calculations. If you run two shifts with different operator skill levels, you need to stratify. The purpose of the index is not to generate a number for the customer's PPAP package. It is to tell you exactly where your process is unstable so you can fix it before it produces a defect.
The same logic applies to OEE. I have seen plants celebrate an OEE of 85% while their availability was 95% and their quality rate was 70%. The aggregate number hid the fact that the process was bleeding defects. Break OEE into its components — availability, performance, quality — and set independent targets. The quality rate component is the one that quality engineering owns. Drive that number, and the rest follows.
Process Capability Targets by Application
Sustaining the System: The FOREAST Principle in Practice
The hardest part of quality management is not the launch. It is the sustainment. Systems degrade because ownership drifts. The production team stops updating the control plan. The engineering team changes a tooling design without triggering a new PFMEA review. The supplier changes a sub-tier process without notification. Each drift is small. Cumulatively, they destroy the system.
Years ago, I developed a sustaining framework based on seven disciplines I applied through my consulting practice: Focus, Organise, Reinforce, Execute, Analyse, Standardise, Transform. The mechanism is straightforward. Focus on the critical few processes. Organise the resources around them. Reinforce through daily gemba walks. Execute the reaction plans. Analyse the data. Standardise what works. Transform the process when the data demands it. Today at a major aerospace manufacturer, these principles are embedded in how I manage routing verification and process governance.
The standardisation step is where most plants fail. A team solves a problem on Line A and never transfers the learning to Line B. The solution dies with the project. Standardisation means updating the master control plan, retraining the operators, and auditing the new standard until it becomes habit. Without that step, you are running a quality improvement programme, not a quality system. The system is what runs when you are not there.