There is a particular silence in a plant when the launch war room realises the defects on the new programme carry the same root causes as the defects on the old one. I have stood in it, usually around week six, when the containment cages meant to vanish at ramp start are multiplying instead. Every struggling OEM reaches for a halo programme to write the turnaround story, while the warranty curve underneath keeps telling the old one – patient as a metronome. Stellantis publicly wrestling quality concerns while betting the recovery on extended-range electric vehicles is the pattern, not the exception. The propulsion is new. The escape paths are not.

Quality debt compounds faster than engineering pride

An open 8D does not pause for a launch date. Supplier PPM keeps its own calendar. Rework loops do not shrink out of respect for a new architecture; they get a fresh failure surface to feed on, with more electronics, more software and more customer scrutiny than anything the plant has shipped before. Quality debt behaves like technical debt in software. Left unpaid it does not level off – it accrues interest in the field and matures twelve to twenty-four months later as warranty.

Every escape shipped is a loan against the brand – the customer pays the interest, and the warranty ledger collects the principal.

I learned that arithmetic at SNOP, building the quality function of a 900+ employee greenfield plant. The 70% defect-cost reduction and the 98% customer satisfaction arrived before we scaled velocity, not after – which is why scaling was survivable at all. The quarter with zero critical customer escalations was not won by heroics. It was QRQC at the line every morning, layered process audits a senior owner actually signed, and 8Ds closed on velocity rather than paperwork. De-escalation taught me the same lesson from the other side: customers calm down when the boring fixes hold.

New propulsion means failure modes your PFMEA library doesn't have

An extended-range EV is not a battery car with an engine in the boot. It is an engine married to a battery, and neither subsystem arrives with a validated marriage certificate. The engine never drives the wheels. It runs as a generator, holding sustained high load instead of transient traction duty – a thermal, lubrication and wear profile your occurrence data has never recorded. Exhaust heat and coolant loops now sit beside a high-voltage pack with its own narrow thermal window. Software decides when the engine wakes, and an EV customer who has never heard an engine in the cabin will log a noise complaint the first time it does. Functional perfection, perceived failure.

None of this sits in the archive. That is the trap.

The bill of materials looks mature, so APQP quietly treats half of it as carry-over. At Witte Automotive I chased fastener failures on parts with decades of clean field history – until a new architecture changed the load path and hardware that had held for fifteen years started loosening in months. The PFMEA occurrence ratings were not wrong; they were answering a different question. When the load path changes, every mature part is a new part, and the FMEA needs re-baselining with new severity, new occurrence and considerably more humility.

The launch starves the base

Here is the quieter mechanism, the one that never reaches the investor deck. The hero programme does not just add workload. It cannibalises the exact people holding the base together: the best containment engineers, the supplier-quality hours, the plant quality leaders – all diverted at the precise moment scrutiny peaks. VDA 6.3 audits get postponed, then quietly cancelled. QRQC shrinks to fifteen compliant minutes. Meanwhile the quiet invoice grows. Third-party sorting at €15,000–€20,000 a week. Premium freight across Europe at €8,000 a shipment. Containment warehouses rented for eight weeks and still standing at week eighty. I have signed premium-freight invoices that would have funded a full PFMEA re-baseline with change left over.

The order of operations is the strategy

Leading quality initiatives across a 2,000+ workforce on multiple sites taught me that sequencing is the job. Not charisma, not dashboards – sequencing. Stabilise the escape rate first: QRQC at the point of occurrence, layered audits with named owners, 8D closure velocity on the production KPI board, A3 discipline for whatever keeps coming back. Then add programme velocity. Launch the other way around and two clocks start together – the ramp and the debt's maturity – so warranty accruals land exactly when the business plan assumed contribution margin.

None of this argues against EREVs; the archetype may be exactly right for the market. It argues for order: a contained base, an honest FMEA re-baseline, supplier readiness proven by audit rather than by assurance.

Key takeaways

  • Escape rate before programme velocity – stabilise with QRQC, layered audits and 8D closure velocity first, or the debt and the bet mature on the same clock.
  • Treat every carry-over part in a new architecture as new: once load paths and duty cycles change, the historical occurrence data describes a part that no longer exists.
  • Budget the quiet invoice – containment, sorting and premium freight, in €, per programme – before launch instead of discovering it in month three.
  • Give APQP and VDA 6.3 gates real veto power; a gate that cannot hold a milestone is documentation theatre, not control.

The uncomfortable observation, then: Stellantis's real decision is not whether to bet on EREVs. It is what gets repaired first. Quality debt compounds, new architectures confound old diagnostics, and a hero programme launched on a leaking base converts ambition into warranty accruals with mechanical reliability. Stabilise the escapes, then launch. A range extender extends range. It has never once extended capability.