I have taken the 3 a.m. call. It never opens with the defect. It opens with a number – a VIN range, a batch code, a shift date – and the slow arithmetic of how many weeks that part has been shipping. Then comes the rented warehouse for quarantined stock, the containment team on weekend overtime, sorters under lamps with gauges, and three weeks of your best engineers doing work that creates value for nobody. Last week, dealer associations in two markets told drivers that frequent recalls are "a commitment to safety, not a reason to panic." The one in South Africa said there was no need to panic at all; its counterpart in Eswatini called recalls proof the system works. That reassurance is written for buyers, and buyers are entitled to it. But plants hear it too, and inside a factory "routine" is the most expensive word in the language.

An apology with logistics

Strip away the regulatory language and a recall is an apology with a supply chain attached. It remediates the customer – replacement part, labour, a letter of goodwill – and corrects nothing on its own. Correcting the process that let the defect through is the 8D's job, if the 8D ever gets the resources, because the campaign team eats the engineers first. I once watched an organisation celebrate closing a field action in record time while the root-cause analysis sat in review for six weeks. We were proud of the apology. Nobody measured the cure.

The economics are unforgiving. A fault that costs €1 to catch at the station costs €10 at the customer's dock and four figures per unit once it reaches the field, before lawyers and brand damage reach the spreadsheet. And the recall is always the last domino: the defect shipped for weeks or months before anyone noticed, which means detection, not recall execution, is where the system actually broke. Recall completion rate is the wrong KPI to celebrate. It measures the ambulance, not the guardrail.

A recall is the only apology in this industry that ships with a tracking number.

How normalisation shows up on the floor

Normalisation does not announce itself. It arrives as vocabulary. Escapes become "campaigns." Recalls become "field actions." Field actions get severity codes and queues, triaged like maintenance tickets, until a weekly meeting exists to manage their flow – a meeting nobody founded with the intention of ending. The language does real work. Every rename moves the event one step further from the word failure and one step closer to the word process.

The paperwork follows the vocabulary. A3s get closed on the strength of "operator retrained" while the PFMEA sits untouched, the same failure mode left fully credible for next quarter. The learning loop is skipped because learning was never the deliverable – closure was. And when a dealer association calls frequency a commitment to safety, the signal dies on the spot: the plant's deviation becomes the industry's baseline. Baselines do not trigger escalations. They trigger budgets.

The quarter with zero escalations

Across two decades in automotive and aerospace, the cleanest evidence I can offer against normalisation is a quarter at SNOP with zero critical customer escalations. Nine hundred employees, greenfield plant, a quality department I built from an empty floor – so I know exactly what that quarter was made of. None of it was luck. QRQC inside 24 hours: any anomaly raised at the station was contained and answered the same day, and a second occurrence opened an A3 automatically. That discipline, not recall logistics, is what over time drove a 70% reduction in defect costs. Prevention has a P&L signature, if you keep the books honestly.

Traceability did the rest. When an escape did occur, records bounded exposure to one shift of serials, so the conversation with the customer was "sort 400 parts in your warehouse," not "campaign the fleet." Earlier, at WITTE Automotive, the Q-Wall gave that principle physical form: suspect stock stopped at a defined boundary, sorted under defined rules, by people who had drilled it while nothing was wrong. We rehearsed containment the way others rehearse fire drills. The failure-cost reduction that followed was large enough that finance stopped treating quality as insurance.

De-escalation, which later became a speciality of mine, is mostly this: never letting a problem grow the legs it needs to reach the customer. The cheapest recall is the one stopped at the wall.

Key takeaways

  • Count recalls as failures, not as responsiveness. Track escapes per million shipped and time-to-detection; campaign completion rate flatters the ambulance and ignores the guardrail.
  • Keep detection inside 24 hours. QRQC at the station, with an automatic A3 on recurrence, kills field actions before they exist.
  • Bound exposure by design. Traceability to one shift of serials turns a fleet campaign into a warehouse sort – build it long before you need it.
  • Drill containment when nothing is wrong. A Q-Wall that has never been rehearsed is decoration, not a control.

So let the dealer associations do their job. A calm driver is a good outcome, and buyers deserve the reassurance. What the plant does with the same words is the real test – the sting belongs inside the fence, not in the press release. The day a recall reads as routine logistics is the day you can be certain the system failed months earlier, quietly, at a station somebody trusted.