The audits multiply. The authority doesn't.

The call always comes late on a Friday. Usually purchasing, sometimes the programme office, once a plant manager on his mobile from an airport departure gate: "The audit report – how serious are the majors, really?" Nobody on that call is asking about the majors. They are asking whether the launch date can outrun the findings. So the news that Tesla is widening its robot-supplier audits across the Yangtze River Delta as volume production approaches did not feel like news. Across two decades in automotive and aerospace I have watched this season repeat on every first-of-a-kind ramp: audit volume grows fastest at exactly the moment the findings become inconvenient.

Here is the uncomfortable observation. This is a non-automotive launch being run with automotive launch discipline, and it is straining the oldest seam in that discipline – the gap between audit coverage and audit authority. On a first-of-a-kind ramp there is no stable process to audit for conformance yet. If the audit cannot judge supplier maturity, and cannot stop a shipment, it is not a control. It is a forecast.

There is no stable process to audit yet

Conformance auditing assumes a frozen baseline: a control plan that matches reality, a PFMEA written by the people who actually run the line, a process that behaved last month the way it behaves today. A robot supply chain months from volume has none of that. Tolerances are still moving. Sub-suppliers are still being swapped. Half the documents on the supplier's wall describe a process that was replaced in August.

I have run enough VDA 6.3 process audits to know which questions predict escapes and which ones decorate the report. The decorating questions are easy: training matrices full of signatures, signage, a tidy cabinet of records. The predictive ones sound mundane. Show me your last deviation and who answered for it – a name, not a department. Tell me what happened on the Friday afternoon when the shift was short and the truck was waiting. Show me whether the corrective action survived the weekend or died quietly when its author went on holiday. A shift leader who can name his top three defects without looking at a screen tells you more about the launch quarter than any score in the audit summary. That is maturity, and maturity is what you audit when conformance has nothing to stand on.

A finding is only as good as its authority

A finding without a stop-ship gate behind it is not a control. It is a forecast with a cover page.

I learned that lesson building the QA/QC department of a greenfield automotive plant – 900+ employees, quality from bare concrete, production starting while the second hall was still a steel frame. On critical part families, supplier governance carried a veto: nothing shipped to the customer without quality release, and quality release meant a signature that was not negotiable. On everything else, governance got a shrug – a report, a rating, a slide in the monthly review. The veto cost us late nights, two furious programme managers and premium freight nobody enjoyed paying. The shrug cost more: sorting teams at the customer, escalations, credibility burned in meetings I could not win. We closed the first quarter of production with zero critical customer escalations, and that quarter was bought by the veto, not by auditing harder. The audit volume was identical either way.

Authority does not need to be dramatic. A gate in the release process. A rule that open majors on safety-characteristic parts freeze the shipment until quality lifts them. A signature that cannot be delegated past a defined severity. That is the whole mechanism. Everything else is documentation.

Containment beats coverage

You cannot audit an immature supply base into maturity before launch. No quantity of audit days achieves it. What you can do is decide, early and coldly, where the firefighting stations go: sorted at the dock, contained at the customer, witness points at the supplier – your engineer standing at the two operations that actually matter, watching the parts that can stop your line.

The muscle memory is QRQC and Q-Wall. At a Tier 1 plant I ran quality for, we cut failure costs substantially by putting every defect on the wall, hour by hour, with a name and a deadline beside each one – QRQC for the fast movers, A3 for the failures that needed a root cause rather than a reaction. The arithmetic did the persuading. A defect caught at the supplier's outbound inspection costs rework and re-sort: cents per part, a few thousand euros a year. The same defect found at your dock costs a containment team, premium freight and a line slowdown: tens of thousands. Found at the customer's line or in the field, it costs 8D hours, expediting, credits and a business review that stays on the record – a hundredfold or worse before anyone counts reputation. Coverage does not change where a defect is found. Containment architecture does.

Expanding audit coverage during a countdown is not discipline arriving. It is risk discovered late. The ramps that survive spend earlier – on supplier development, on containment architecture, on giving findings teeth – so the launch date never has to race the audit calendar at all. Left to themselves, the audit calendar always loses to the launch calendar. Only the veto makes it a contest.