A customer takes delivery of a new EV. Inside four months the charge controller has been flashed twice, a HV harness connector has been replaced under goodwill, and the cabin carries a faint coolant smell when the pack preconditions in winter. I have run enough lines to know what that combination means. Three release gates, a full PPAP package, a passed IATF surveillance audit — all cleared, none of it catching the failure modes that actually matter. The coverage this week on Kia and Hyundai owners hitting exactly this kind of trouble is being read as a Korean problem. It isn't. It is the predictable outcome of running a combustion-era quality system against a product whose dominant failure modes were never in the FMEA.

Your PFMEA was written for an engine that isn't in the car

After a century of internal combustion, the failure surface is mapped. Bearing wear, oil degradation, valve-seat recession, gasket blow-by, ring sticking — the process data reaches back decades, and the RPNs in your PFMEA reflect that inheritance. You know what severity to assign a connecting-rod failure because the plant your parts came from has known for forty years.

Open the EV equivalent and the filing cabinet thins out fast. Thermal propagation through a cell stack after a single cell vents. Inverter firmware drift following an OTA update nobody revalidated against the hardware-in-the-loop bench. HV connector fretting under vibration profiles the supplier never received, because the OEM trimmed the road-load file to fit an email attachment. These are not edge cases. They are the dominant modes of the architecture, and the warranty department is finding them — not the design FMEA.

I have sat in enough APQP reviews to recognise the look on an engineer's face when a failure mode is genuinely new. The PFMEA gets a row added after the fact. RPN nudged upward, action assigned, document re-issued. Everyone moves on. The file becomes a memorial to what already failed rather than a predictive instrument. IATF 16949, read honestly, demands the opposite — PFMEA is a living engineering tool, re-opened at every design change and every process drift. At WITTE Automotive we used QRQC and A3 to drive substantial failure-cost reduction because we treated those documents as working tools, not archived evidence. Most organisations never build that habit. The document sits. The failure repeats.

A PFMEA that has never predicted a real failure is not a quality tool. It is a liability shield with a table format.

The supplier scorecard assumes a maturity that doesn't exist

Run a tier-1 supplier audit on a brake-caliper manufacturer and you are leaning on three decades of process capability data. Cpk on the bore, SPC on the surface finish, a PPAP that has been re-levelled across two generations of the same part. The scorecard works because the underlying process knowledge is mature.

Run the same scorecard against a battery cell supplier founded in 2021 and the spreadsheet becomes theatre. You get a capability study on three pilot-build batches, a cleanliness specification copied from a competitor's datasheet, and a quality manual that reads better than the cells perform. The maturity gap between a tier-1 with thirty years on the clock and a cell maker on its second serial programme is not a few points on a rating. It is the difference between governing a known process and gambling on an unknown one. The standard supplier quality framework — IATF, VDA 6.3, the layered audit cadence — was calibrated against mature combustion-era supply chains. Apply it unchanged to an EV bill of materials and you are scoring against a curve the data does not support. The audit passes. The cells still drift.

You revalidated the part. You didn't requalify the process.

This is the gap nobody on the launch deck wants to name. PPAP covers dimensional and material conformance at launch volume. It does not cover the knowledge vacuum that opens when every sub-assembly is new and the plant has zero tribal memory of how it fails. A revalidation says the geometry is right. It says nothing about whether the organisation knows the part.

I lived the inverse of this when I built the greenfield QA/QC function for 900-plus employees at SNOP. No inherited assumptions. No legacy FMEA to lean on, no old-timer who remembered the 2014 batch. Every process failure mode had to be discovered, mapped, and engineered out from first principles. We took that posture to 70 per cent defect-cost reduction and 98 per cent customer satisfaction inside the programme window. The lesson was unambiguous: a quality system is either constructed for the product it governs, or it is a costume.

Most OEMs entering the EV space have done the cosmetic work. The badges are electric. The quality system underneath is still wearing last decade's clothes.

Key takeaways

  • An inherited PFMEA is the single biggest liability on an EV programme — audit it line by line against actual field failures, not against the previous model year's filing.
  • Supplier scorecards need a separate track for components whose underlying process is under five years old; weight the capability data accordingly and stop pretending the Cpk is stable when the process that produced it isn't.
  • PPAP revalidation does not equal process requalification — tribal knowledge is a launch input, and its absence is a risk to be managed, not assumed away by a document package.
  • Greenfield discipline — building the quality system from zero rather than inheriting it — is the only honest posture for an EV programme that intends to survive warranty year three.

The OEMs that come through the EV transition intact will not be the ones with the best cells or the most polished launch deck. They will be the ones who looked at their own quality system, admitted it was built for a product they no longer make, and started over. Same as any greenfield. Same as the day the line was empty and the playbook had not been written yet.