Your laser scanner measures to the micron. It can't catch a data typo.

I've stood in enough boardrooms watching executives nod along to million-euro inspection equipment proposals to know how this story ends. The multisensor measurement market is booming — laser scanners, vision systems, CMMs with sub-micron capability — and manufacturers keep writing the cheques. Meanwhile, Subaru just recalled 541,237 Crosstreks, Foresters, and Ascents. Not because an axle fractured. Not because a weld failed. Someone entered the wrong Gross Axle Weight Rating into a document and it propagated through the chain unchecked.

A laser scanner reading to the micron cannot catch a data entry error. And nobody in the quality review meeting seems to find that worth discussing.

Inspection validates the part, not the paperwork

Here is an uncomfortable observation from two decades of running plants: we build elaborate inspection regimes to verify that physical parts match specifications, then quietly assume the specifications themselves are correct. A multisensor CMM will confirm that an axle housing was machined to ±0.005 mm. It cannot tell you whether the tolerance value in the release documentation matches the engineering drawing, the ERP master data, and the regulatory certification filing. That chain is invisible to the scanner.

The Subaru case is textbook. The axles were manufactured to print. The GAWR figure — a regulatory and safety parameter — was wrong in the documentation. Half a million vehicles pulled back over paperwork. Not a single CMM in the supply chain would have flagged it, because none of them is designed to. They inspect hardware. They do not audit data lineage.

Your database is a critical-to-quality process

I have walked through plants where APQP and PFMEA rigorously govern every physical operation — stamping, welding, machining, torque application — while the label printer in shipping and the ERP update workflow sit completely outside the quality management system. That is a structural failure dressed up as normality.

When a tolerance change hits engineering but the change control doesn't propagate to the production work instructions, the supplier PPAP records, and the customer-facing certification in the same controlled cycle, you have a data integrity nonconformance. It is as real as a dimensional deviation. It reaches the customer just as fast — often faster, because nobody is looking for it.

Every system that carries product specifications — PLM, ERP, MES, label databases, regulatory filings — should be treated as a critical-to-quality process. That means PFMEA on the data flow, control plans on the change points, verification at each handoff. If that sounds excessive, price it against recalling half a million vehicles.

What routing verification actually delivers

At Airbus, I led the implementation of Routing Verification KPIs across the manufacturing engineering organisation in North America. The objective was not another layer of automated inspection. It was to validate the integrity of the data flowing between systems — engineering releases to work instructions, work instructions to tooling definitions, tooling definitions to measurement plans. Each routing point received a verification gate and a measurable KPI.

The result was a 97% reduction in internal lead time. Not because we scanned more parts or added another vision system. Because we stopped discovering data errors downstream — at assembly, at audit, at customer acceptance. When the data flow is clean, the physical process follows. When it is broken, no volume of inspection compensates.

That 97% came from treating data integrity as a manufacturing process and governing it with the same discipline applied to torque values or weld penetration. The inspection equipment validated the outcomes. It did not produce them.

Stop buying hardware to solve a systems problem

I have reviewed too many capex proposals that frame another automated inspection platform as the answer to defect-cost reduction. Manufacturers buy the hardware, commission it, then discover the next recall originated from a clerical error in a weight rating.

The most expensive defect in your plant is the one your inspection system was never designed to find.

QRQC, A3, 8D — the backbone tools of operational quality — exist for systematic root-cause identification and prevention, not for detection theatre. If your root-cause analyses keep landing on documentation discrepancies, specification mismatches, or data entry errors, the fault is upstream in your governance. Another scanner will not touch it. At WITTE Automotive, the substantial failure-cost reductions I delivered came from Q-Wall discipline and structured QRQC responses, not from adding measurement stations. At SNOP, where I built the quality function for a 900-person greenfield plant from a bare concrete floor, the 70% defect-cost reduction we achieved was rooted in process control and data discipline long before anyone spec'd a vision system.

The Subaru recall will cost millions across logistics, dealer mobilisation, and reputational repair. It will not trace back to a machining cell or a faulty robot. It will trace back to a number, typed somewhere, that nobody verified because nobody owned that handoff. Your next quality investment should not be another scanner with finer resolution. It should be the discipline to govern the data that tells every device in your plant what to measure.