You can integrate the evidence. You can't integrate the verdict.

Here is an uncomfortable observation. Every aerospace ramp in living memory has been followed, with metronomic regularity, by a proposal to "integrate" quality. Quality Magazine ran the question again this week – when the industry will finally trade independent quality management for "integrated assurance ecosystems" – and the timing tells you most of what you need to know. Orders are pouring in. Lead times are compressing. Every function that does not ship parts is being asked to justify its headcount.

I have met this proposal before, under three different names, and it has always protected the same budget line. The word ecosystem has never appeared on a nonconformance report. An NCR deals in part numbers, heat lots, torque values and signatures – things that are allowed to fail. So the real question is not whether to integrate. It is which half of it you are being sold.

What integration should actually buy you

The honest half is connected evidence, and it is overdue. Supplier first-article data, SPC charts, audit findings, NCR dispositions, 8D actions, containment records – all of it belongs in one traceable thread, so that when a batch turns suspect an engineer rebuilds its history in minutes instead of losing half a week to archaeology across the QMS, the MES, a supplier portal and somebody's shared drive. I have watched that archaeology at plants on both sides of the automotive–aerospace fence. It is a self-inflicted wound. And EN 9100 already demands the traceability – you would be integrating work the standard requires of you anyway.

Connected evidence also sharpens verdicts. An MRB that sees supplier PPM, process capability and prior escape history in one view dispositions faster and more defensibly. Connect the data. Nobody serious argues against that half.

Where independence is load-bearing

The trouble starts with the other half. The integration deck is not about data plumbing; it means merged reporting lines – quality dissolved into operations, one org chart, one set of priorities, one throat to choke. Watch what happens to the four verdict rights: disposition authority, MRB, stop-ship, supplier escalation. The moment the person who dispositions a nonconformance reports to the person who owns the delivery schedule, your PPM becomes a schedule KPI in disguise.

Nobody falsifies anything. The system simply makes "no" expensive. The use-as-is call on a €40,000 machined housing drifts from engineering judgement to throughput arithmetic, one meeting at a time.

Independence here is not bureaucracy. It is structure. The aerospace standard family assumes that whoever verifies conformity is not the person rewarded for producing it; IATF 16949 lets you blur that line with competence and process controls, and automotive plants do – with results I will come to. A quality function that can be overruled for free will be overruled whenever the month is tight. Escapes are manufactured in precisely that gap. They are the most expensive parts you will ever ship.

A quality verdict is worth exactly what it costs to overrule it.

Automotive already ran this experiment

Two decades across automotive and aerospace – the CV reads Airbus, Witte Automotive, SNOP – have shown me both standards from the inside. IATF plants embedded quality engineers in operations twenty years ago. Some sites kept independent verdict rights: disposition, containment, stop-ship. Some handed them to production in the name of flow. The pattern held wherever I looked. Sites that kept the verdict independent kept their escapes down. Sites that gave it away found their customers were running the MRB for them, at considerably higher cost.

When I built the QA/QC function for a greenfield plant of 900+ employees – nothing existed, not one inspection instruction – I spent more time on reporting lines and decision rights than on CMM selection. Before the first part shipped, it was written down and signed: disposition authority sat with quality, the MRB was chaired by quality, production could escalate a verdict upward but never overrule it sideways, and stop-ship required nobody's permission on a Friday afternoon. Those org-chart choices surfaced later as numbers. Zero critical customer escalations within a quarter. A 70% defect-cost reduction. A function with real authority says no early, when no is still cheap – and QRQC only runs at full speed where the verdict stands. Most escalations I have had to defuse in my career trace back to a verdict that was overridden, or never independently made at all.

Independence is not overhead. It is the cheapest escape insurance on the market.

So independent versus integrated is a false binary sold as modernisation. Integrate the evidence. Keep the verdict separate. Be honest about which of the two the budget line is aiming at. The demand surge will pass, the backlog will be delivered, and the escapes manufactured in the meantime will be paid for across a decade of 8Ds and customer audits. If your assurance ecosystem contains no node that can say no, it is not assurance. It is documentation with better UX.