This week Stellantis is promising, once again, to sort out its quality, and even the motoring press filed the story with a weary aside: we have heard this before. My reaction is less cynical than practical. Nobody can check that promise for two model years. I have run a real turnaround at plant scale, and the announcement was the easiest part by a distance – an afternoon and a press release. The turnaround itself was 06:45 QRQC meetings at the line and a year spent shortening the distance between a defect appearing and a defect being contained. I have also stood in a quality review where the only honest object in the room was a bin of rejected parts with the customer's logo moulded into the plastic. That bin created more urgency than any commitment we ever announced.
Why the announcement is free
Warranty truth has a clock, and the clock is slow. The industry grades itself in 12MIS and 36MIS – incidents per thousand vehicles at twelve and thirty-six months in service. A car launched last quarter reveals its warranty character somewhere between next year and the year after. The recalls filling this month's headlines were designed, sourced and launched two to three years ago. So an executive promising better quality today has made a claim the field population cannot begin to refute for twelve months. Cannot fully refute for thirty-six.
Leadership can outrun a news cycle. Nobody outruns a warranty cycle.
None of this requires bad faith. It is structural. The announcement rides an evidence vacuum for exactly as long as the news cycle needs it to, and nobody outside the warranty department can audit the difference between a changed company and a changed message.
What a real turnaround changes by Tuesday
The test I applied in the plants I was accountable for, and later at suppliers through consulting work, was simple: if a turnaround is real, the floor looks different within a week. Not the five-year plan – the floor. QRQC runs daily, at the line, within twenty-four hours of detection, with the operator who found the defect standing next to the engineer who has to close it. Detection-to-containment is measured in hours, because that is the only interval that moves before everything else does. Reaction plans carry a name and a date – not a department, a person who gets called on Sunday. And stop-the-line authority gets used, with the person who pulls it thanked in front of the shift rather than coached about overreacting.
At SNOP, a 900-plus employee greenfield plant where I built the QA/QC function from the ground up, defect costs fell 70%. That is not a communications artefact. It is the accounting department's number. It fell because the loop shortened – failures surfaced hours after they occurred instead of weeks, and were contained at the station instead of leaving as a customer's freight claim. Customer satisfaction at 98% was a consequence, not a slide.
At Witte Automotive the mechanism was the same, built on QRQC, A3 and the Q-Wall. The Q-Wall did one unglamorous thing: it put the failure data on a wall in the path of the people who could act on it, exactly where a promise would normally have hung. Failure costs came down accordingly. Nothing improved because it was announced. It improved because the distance between truth and action shrank.
How customers decide whether to believe you
I spent a stretch of my career as the man sent in when a demanding OEM customer had stopped being polite. Within one quarter I brought a multi-site operation to zero critical customer escalations. The lesson transfers to any audience: credibility is not bought with commitments. It is bought with evidence velocity.
Every serious customer keeps a private ledger. It records the speed of your disclosure against the speed of their discovery; whether your 8D named a physical root cause or offered "operator awareness", which every SQE on earth reads as "we do not know"; whether the control plan lies open on the table and the first-time-through trend arrives before anyone asks for it. A supplier who phones the customer before the customer's own field data does keeps the relationship. One who waits to be discovered does not. Customers forgive defects. They do not forgive slow truth.
A three-question test for any quality promise
So when an OEM announces a quality turnaround, I do not ask whether they mean it. I ask three questions, and they work on any promise in any industry:
- Name the changed process step. Not the department, not the initiative – the station, the gate, the reaction plan that behaves differently this week than last.
- Name the owner. One person, with a phone number. If the answer is a steering committee, nobody owns it.
- Name the leading indicator that moves within ninety days. QRQC closure time, first-time-through, containment escapes, 8D cycle time – something that will move long before 12MIS can testify.
Key takeaways
- Warranty truth lags launch by 12–36 months, so any turnaround claim younger than that is a hypothesis, not a result.
- Judge a turnaround by Tuesday, not by the model year: QRQC cadence, detection-to-containment in hours, reaction plans with named owners, stop-the-line authority that gets exercised.
- Customer credibility is bought with evidence velocity – disclosure speed, root-caused 8Ds, an open control plan, a first-time-through trend shared unprompted.
- Run the three-question test: changed process step, named owner, leading indicator inside ninety days. A reorganisation or a dashboard fails it.
If the honest answer to those three questions is a reorganisation or a dashboard, you have heard this promise before and you will hear it again. None of this proves Stellantis is insincere. It proves nobody can know yet, and the structure of the promise protects the promiser. But the plant knows by Tuesday. The customer's ledger knows within a fortnight. The field population renders its verdict on schedule, at twelve and thirty-six months, indifferent to everything that was announced. The only question worth asking is whether anyone changed anything on the floor in time to earn a different answer.