The design is mature. The plant is day one.

Zaragoza has started building the Leapmotor B10 – a European twin of a Chinese plant that has run the same design long enough to have forgotten its own teething troubles. In the same week, the export press published its Chinese-EV reality check, and sceptical customers are grading panel gaps and software with a jeweller's loupe. Here is an uncomfortable observation from two decades of launches: a mature design entering a new plant does not behave like a mature product. It behaves like an immature one in a borrowed suit. The paperwork transfers. The scar tissue stays home.

What a twin plant actually inherits

The transferable layer of a plant is thin and travels fast: CAD, cycle times, control plans, work instructions, supplier lists, tooling drawings. A committed programme office moves all of it across a border in weeks. What stays home is thicker. The torque spec says 47 Nm. It does not say that 46 loosened after thermal cycling, and that somebody found out the expensive way during a humid August. Nobody writes "we shimmed that jig in March and it never troubled us again" into a control plan. The parameters arrive without their reasons.

This is why I tell clients to read a transferred PFMEA as a diary of someone else's pain. You inherit the severity–occurrence scores – the conclusions – without the experiments that earned them. A Cpk from the mother plant is a fact about that plant's hands, its maintenance culture and its climate. Not about yours.

Believe the design. Withhold belief from the capability numbers until your own line has produced them.

You can email a control plan across a border in seconds. The flinch takes years.

Day one has its own failure modes

New hands come first. A defect an operator has never seen is invisible, and detection is a skill built by escapes – some of which reach the customer first. New maintenance culture follows: the same press with a different lubrication schedule and a different reflex when an alarm sounds is not the same press. Zaragoza's dry heat against a humid Yangtze-delta summer shifts adhesive cure windows, static behaviour and sealing performance in ways no work instruction anticipates. And localised suppliers are qualifying in real time, re-running PPAP on parts the line is already consuming.

I lived the pure version of this. At SNOP I built the quality organisation for a greenfield plant of more than 900 people – every process existed on paper, none had memory. The first months were a merciless QRQC cadence: problems captured at the station, closed at the board, escalated the moment containment failed. The 70% defect-cost reduction we eventually delivered, on the road to 98% customer satisfaction, did not arrive because the documents were good. It arrived as scar tissue accumulated, one closed 8D at a time. And the quarter with zero critical customer escalations was bought, expensively, up front – extra verification, conservative boundaries, layered checks at the gate – during the ramp, when maturity did not yet exist to be relied upon.

How to buy maturity back

Maturity cannot be shipped. It can be regrown, deliberately, and the mechanisms are old ones. Put mother-plant veterans on the line, not in the office: end-of-line plus the three historically worst operations, on shifts, with translation. Their value is not process knowledge, because that travels in the documents. Their value is the flinch – the half-second hesitation before a nuisance becomes a defect.

Keep golden units as arbitration masters: one known-good, one known-bad, retained from the mother plant and flown in before SOP. Every borderline call by a new inspector, every measurement dispute, settles against them in minutes instead of in an email thread that outlives the argument. Run PFMEA reconciliation workshops where the local team defends each high-severity failure mode in its own words – cause, current control, detection – and whatever cannot be explained gets a containment until it can be. And requalify local suppliers on their own data. "Same part number, same PPAP" is a claim about a drawing, not about a process; capability lives in machines and hands, and both changed address.

Then taper the containment on evidence, not on anniversaries. At WITTE we ran the Q-Wall – layered end-of-line verification reviewed daily – and a check came off when escape rates and detection latency proved the process held without it. A containment retired on a date is a bet. Retired on data, it is a decision. Containment is not free, either: an extra verifier per shift is €50,000–60,000 a year. Still the cheapest tuition on the market – the alternative is a customer teaching your plant the same lesson at retail price.

The twin-plant model is sound. It removes most of the product risk a ground-up launch carries, which is exactly why it is seductive. But no plant was ever copied – only rebuilt – and the rebuild runs at the speed of accumulated judgement, not at the speed of the transfer budget. SOP is a supply-chain milestone. Maturity is earned on the curve, shift by shift. And the customers who spent this week publishing their Chinese-EV reality check will grade the first two quarters hardest, because they understand what they are watching: a mature design meeting a day-one plant. They know which of the two they are buying.