A Tier 1 automotive supplier ships a batch of 240 parts to a German OEM. Four weeks later, the customer reports incorrect surface hardness. The claim costs €187,000. The 8D investigation reveals the defect could have been caught at five distinct points in the process.
Five opportunities to catch the error. Zero successes. I have audited plants across automotive and aerospace for over twenty years, and the root cause is almost always the same. Nobody ever sat down and mapped their barriers on a single sheet of paper.
Barrier Analysis originated in health, safety, and environmental management. Applied to IATF 16949 and AS9100 quality systems, it is a method for seeing every control, every mechanism, and every safeguard between a potential failure and the customer—and critically, identifying where the holes are.
The Anatomy of a Failed Defence
Take the hardness defect. The supplier’s cross-functional team finally mapped the process from the heat treatment furnace to the customer's incoming inspection. They found five barriers, and every single one had a structural weakness.
Barrier 1 was furnace calibration. It existed but was passive—a six-month check that could not detect a 12°C sensor drift occurring in real time. Barrier 2 was the temperature alarm. The threshold was set at ±15°C. The 12°C drift slipped right through.
Barrier 3 was the first-piece inspection. A technician took samples, but on the day of the failure, a substitute recorded the results against the wrong batch ID. Barrier 4 was the final QC check—a 2% random sample. Because the drift was systematic based on furnace position, random sampling from the same position missed it entirely. Barrier 5 was the OEM's incoming inspection, which only checked every 50th shipment.
Five barriers. Five blind spots. No single mechanism was independently capable of catching the specific failure mode.
Barrier Vulnerability in Practice
What the team assumed
- Six-month calibration prevents drift
- The ±15°C alarm threshold is tight enough
- First-piece inspection catches batch errors
- 2% random sampling reflects the whole batch
What actually happens
- Drift happens continuously between checks
- Systemic drift stays below the alarm limit
- Sampling misses systematic positional defects
- Random selection cannot catch systematic failures
The Four Categories of Barriers
To map your defences accurately, you must classify them. A robust process uses a mix of all four. If your barriers are concentrated in just one or two categories, a single systemic event can disable them all.
- Physical: Poka-yoke devices, hard fixtures, mistake-proofing design features that mechanically prevent the error.
- Technological: SPC alarms, automated jidoka line stops, MES rules that block progression based on sensor data.
- Procedural: Control plans, standardized work instructions, flow charts, and defined escalation paths.
- Human: Verified operator competence, clear standards, the authority to stop the line, and a culture of reporting.
Listing these barriers is not enough. You must understand how they connect, where they overlap, and what happens when two fail under the exact same conditions.
Building the Barrier Map
A Barrier Map takes the outputs of your PFMEA and applies them to reality. It traces the path from the error source to the customer, plotting every safeguard along the way. The goal is to evaluate independence and detection capability.
The Barrier Mapping Sequence
- 01Define the scenarioBe specific. 'Hardness failure on part XYZ shipped to OEM in week 22' drives better analysis than 'bad surface'.
- 02Identify the error sourceUse your PFMEA, Ishikawa diagram, or 5 Why analysis to pinpoint where the failure originates.
- 03Map existing barriersWalk the gemba. Trace the process physically from the source to the dock. Log every check, sensor, and manual inspection.
- 04Evaluate independenceDetermine if barriers are active or passive. Assess if they rely on the same operator, machine, or data stream.
- 05Close the gapsAdd new barriers, tighten thresholds, or remove single points of failure. Prioritise by defect prevention and cost.
I ran this exact exercise with a plant in Eastern Slovakia producing engine belts. The scenario was insufficient splice strength leading to field failure in the customer's transmission. The team mapped five barriers between the vulcanization press and the OEM.
Three of the five barriers were completely irrelevant to splice strength. The visual check looked for surface defects. The customer's incoming inspection only checked dimensions. Between the pressure sensor alarm and the visual check, there was a total void. No mechanism existed to catch the defect.
The team installed continuous pressure monitoring with a tight tolerance and automatic press cutoff. They stratified the destructive pull-test sampling across the batch. Implementation cost was €23,000. They saw zero splice-related claims over the next year.
Barrier Analysis takes FMEA outputs and adds physical reality: how your defences are arranged in space and time.
Avoiding the Barrier Illusion
The most common trap is the illusion of protection. Teams claim they have a barrier because an SPC chart exists on a parameter. But if nobody reviews the chart until after the batch ships, it is not a barrier. It is a historical record.
Another fatal error is dependency. If your first-piece inspection, SPC data entry, and final line clearance all rely on the same uncalibrated sensor, you do not have three barriers. You have one barrier three times. A single systemic failure defeats them all.
| Barrier Claim | Operational Reality | Required Fix |
|---|---|---|
| Operator visually inspects | No time allotted, standard unclear | Add physical poka-yoke or sensor |
| SPC monitors the parameter | Data reviewed post-shipment | Automate MES hold based on live SPC |
| Calibration is current | Drift occurs between 6-month checks | Implement continuous monitoring |
Furthermore, treat human barriers with strict scrutiny. "The operator will check" is a hope, not a control. A real human barrier requires verified competence, a clear visual standard, adequate time, and the explicit authority to stop the line.
Implementation and Standardisation
Run a Barrier Analysis after any critical customer complaint to understand how the defect escaped. Use it proactively during PPAP and Run@Rate for new processes. Anytime you introduce a new machine, a new supplier, or a new operator, your barrier map changes.
At the Slovakian supplier from the opening example, management implemented continuous furnace profiling with a ±5°C drift alarm. They locked batch IDs to measurement results in the MES, preventing misidentification. They stratified the sampling plan across furnace positions and instituted an automatic hold if results were missing.
The implementation cost €34,000. It saved over €200,000 in annual claim costs. More importantly, the OEM saw a systematic, engineered defence and renewed their confidence in the supplier.
Barrier Analysis is not about adding more inspections. It is about understanding how your existing controls function together. Once you see the gaps, you can finally build walls that keep defects from reaching your customer.