The recall is 600. The failure mode is 8,878.

Here is an uncomfortable observation from the receiving end of those calls: when the same car, in the same market, goes from 600 recalled units to 8,878 in two years, the failure mode did not grow. The metal did not change. What changed was the size of the honest answer – and somebody is now paying the difference between the first number and the second. Chery's Omoda 5 in Malaysia reads, in the headlines, like a mystery. To anyone who has scoped containment under time pressure, it reads like a receipt.

I spent years as the quality director customers rang first. The opening question from an OEM is always the same three-part litany: how many, where, since when. Not "what happened" – they assume you will get to that. The quality of the first answer decides whether the conversation stays technical or turns commercial. I once went a full quarter with zero critical customer escalations, and the discipline behind that number was not eloquence. It was the ability to answer the litany inside an hour, with lot numbers instead of adjectives.

The first number is the traceable number

A first recall scope gets drawn around what the evidence can defend. Lot boundaries. Changeover records. The supplier batch window your ERP says the suspect parts occupied. If the genealogy is clean, the box hugs the true population. If it is patchy, the box hugs the patch – and a box drawn around your records is not the same animal as a box drawn around the problem.

Then the politics arrive. Warranty accruals need a number this quarter, so smaller is cheaper. The plant manager wants the scope "defensible", which in practice means small enough to survive the morning meeting. Finance wants it bounded; sales wants it invisible. Under-scoping is rarely an accident of analysis. It is a bet that the field data stays quiet long enough for everyone to forget who placed it.

A small recall scope is not a verdict; it is a hypothesis, and the field is under no obligation to confirm it.

How 600 becomes 8,878

Run the arithmetic nobody runs at 2 a.m. One supplier lot crosses three months of builds because the supplier shipped a single batch and your receiving dock never separated it. A process window drifted, so the parts at the edge came out marginal rather than defective – and marginal parts pass gates for months. A part number shared across trims means the suspect population was never one variant. Multiply any two of those and 600 becomes 8,878 without a single new failure mode appearing. The re-scope is not a new problem. It is the old one being revealed, one warranty claim at a time.

Field data grades your first guess, always late – typically eighteen months late, right around the time the second headline drops. Volkswagen has been living a version of this with battery recalls that keep widening; the pattern belongs to no badge in particular.

Escapes are bounded by your records, not your hopes. The physics of the failure does not care what your ERP could prove in 2024.

Build the answer before the question

I learned this the cheap way, on a greenfield site: no legacy systems, no inherited excuses. Building a quality organisation for more than 900 people from the ground up meant we could design lot genealogy into the process from day one – every changeover captured, every supplier lot boundary sitting in the ERP where a query could reach it, traceability priced as insurance in euros rather than filed as paperwork for the IATF 16949 audit. The standard asks you to identify and trace. Auditors check the procedure. Customers check the clock. Traceability that lives in the system rather than in the manual comes down to one thing: when the phone rings, the answer is a query, not archaeology.

The payoff showed up in numbers we could defend. Defect costs came down 70% on that build, and QRQC discipline did much of the work – its quietest contribution being not faster fixes but knowing exactly which population a suspect lot had touched. Containment that takes 40 minutes instead of four days stops suspect parts while the question is still cheap. It also spares you from guessing wide "to be safe", and guessing wide carries its own invoice: bloated warranty reserves and customer noise you later have to walk back.

The invoice for the first number's optimism

The second recall is the invoice for the first one's optimism, and it arrives itemised: reverse logistics, dealer capacity, regulatory re-notification, a second press cycle, and the one line nobody prices – credibility. Customers forgive a big honest number faster than a small dishonest one, because the big honest number says we know our build. A number that keeps growing says the opposite, in public, on a schedule you do not control.

The Omoda arithmetic is not a Chinese problem or a Malaysian problem; it is the standard behaviour of traceability gaps everywhere, and the deferred invoice always clears. The car did not change – the size of the honest answer did. Nobody applauds the person who scopes a recall small. They only delay the applause by eighteen months, and add interest.