Here is an uncomfortable observation from two decades in automotive and aerospace plants: the mark is the first record in your traceability chain and the last process anyone qualifies. Recently a marking-equipment vendor published a note on the cost and traceability risks of poor part marking, and the trade coverage filed it under metrology trivia – a niche vendor, a niche problem. I read it as a confession the whole industry could sign. Lose the data matrix on the part and your ERP becomes a library with no catalogue: every record still exists, none of them can find the book. Traceability lives in the database, as everyone likes to say. The proof is on the part. Those are not the same asset, and only one of them survives the paint shop.

The mark is a measurement process nobody gauges

Watch what happens at launch. Marking gets a specification – AS9132 for dot-peen data matrix, an AIM DPM grade threshold for the verification reader – and then the process is never spoken of again. No capability study on the marker as parameters drift and tooling wears. No MSA on the reader, which is a gauge in every sense that matters and the only gauge in the plant exempt from measurement-system analysis. Audit status: none. Nobody asks whether the reader can repeat its own grade on the same part twice in one afternoon.

The PFMEA tells the same story. Marking appears as a production step – apply code, verify grade, pack – with severity borrowed from a cosmetic defect. The actual failure mode is not "mark looks poor". It is "recall cannot be scoped", a severity-10 event that fires years after launch, in someone else's plant, on parts nobody can attribute. The FMEA never connects those two facts.

When I built the quality architecture for the SNOP greenfield plant – 900+ employees, nothing inherited, nothing grandfathered – marking got the same treatment as measurement. Mark placement relative to forming, verification point, re-grade after coating, and a drift rule: when the grade slipped, escalation through QRQC within the hour, not a quiet tweak of camera gain at the line. The uncomfortable discovery was that mark placement is a process-design decision, because every downstream deformation is a degradation step. Most plants choose mark placement for cycle time. We chose it for the tenth year of service life.

Containment is only as narrow as your readability

At WITTE Automotive I ran the QRQC discipline that cut our failure costs hard, and the lesson that stayed was not a defect – it was a quarantine. A suspect lot surfaced in a latch component. Batch records were clean. The marks were not: the data matrix had been applied before a forming operation, and forming distorted the grid until the reader grade collapsed. At assembly, parts from that run sat in the same bins as the good runs either side of it, and we could not prove which was which.

The initial containment was one shift's output. Without readable marks the net widened to five days of shipments – sorting overtime, customer notification, locked stock. Call it €60,000. The marking process cost a fraction of a cent per part. Recall scope is set on the shop floor, not by the lawyer; the lawyer can only narrow it as far as the evidence allows, and the evidence was unreadable.

Marks must outlive the warranty, not the shift

Your verification camera sees the mark for two seconds at the marking station. Then the part goes through cataphoretic dip, paint, riveting, heat, a decade of salt and service. Every one of those steps attacks contrast, dot roundness and the quiet zone – which is exactly why AS9132 and AIM grading exist as graded scales rather than pass/fail stickers. Standards writers assumed marks degrade. Plants assume they don't.

Aerospace learned this the hard way. Under AS9100 and EN 9100, an unreadable serial on an individually traced part means no proof of conformity at all – there is no lot fallback in a serial-traceability regime, just a bin of anonymous hardware and a paper trail pointing at nothing. IATF 16949 practice is softer, and the softness lulls plants: lot traceability reassures you that the paperwork survives even if the mark dies, right up to the moment you must place one suspect part inside or outside a shipped lot. At that moment lot traceability degrades to "the lot, or everything the lot touched". Automotive warranty departments are relearning this now, one unreadable field return at a time.

A database can only recall what the part can still prove – everything else is a well-documented rumour.

Key takeaways

  • Run MSA on the verification reader and calibrate it like any other gauge – it gates the existence of traceability, so treat its grade as a controlled characteristic.
  • Put marking in the PFMEA as a failure mode in its own right, with severity tied to recall scope, not as a production step carrying cosmetic severity.
  • Re-grade marks after the harshest downstream step – forming, coating, heat – on finished parts, not only at the marking station where the mark is at its youngest.
  • Give marking audit status: a capability study at launch, periodic re-grading with records, and a named escalation path when grade drifts, visible in VDA 6.3 and IATF 16949 walks.

Treat marking as a controlled process – capable, verified, re-graded, audited – and it stays the cheapest insurance in the plant. Skip it, and the first line of your traceability chain is written in a font nobody can read. The record printed on the part is the only one a customer can hold up in a sorting cell, the only one an investigator can hold up in a courtroom, and the only one you cannot restore from backup.