Here is an uncomfortable observation from twenty years on both sides of the audit table: regulators don't count problems. They count symptoms. This week gave us a clean demonstration. Caplin Point's sterile facility drew an OAI classification from the US FDA on the back of ten observations. Days earlier, a unit of Aurobindo Pharma walked away with a VAI on three. Ten versus three looks like arithmetic. It isn't. An OAI is not a defect count — it is a trust verdict. It says the regulator no longer believes your system can self-correct. I have watched that verdict land on automotive and aerospace suppliers wearing different clothes, and the mechanics of climbing back down are identical everywhere.

What a classification actually measures

For readers who live in IATF 16949 and AS9100 rather than pharma: VAI — voluntary action indicated — means the regulator read your responses and believes you. OAI — official action indicated — means they don't. The observation count is nearly irrelevant to the classification. Three findings with evasive responses can draw an OAI; ten with credible corrective logic can, occasionally, avoid it. What gets measured is the auditor's confidence in your management system, not the inventory of your defects.

Anyone who has run plants knows the parallel ladder by heart. It starts quietly, with a customer scorecard going amber. Then controlled shipping level 1 — you inspect your own product under their rules. Level 2 is where a third party stands in your shipping lane and you pay for the privilege. Above that sits the new business hold: your commercial organisation can quote, but nothing gets awarded until quality releases the freeze. In aerospace the same ladder runs through supplier surveillance escalations. Every rung carries the message the FDA sends with an OAI: we no longer trust your system to catch its own failures, so we will externalise the control you failed to internalise.

Ten findings at one site is not ten failures. It is one failure wearing ten coats.

Observations cluster — read them as one tree

The first thing I do when a wall of findings arrives is stop reading it as a list and start reading it as a tree. At WITTE Automotive I inherited an audit report that read like a shotgun blast — documentation, traceability, reaction plans, training, calibration. My team printed every finding on a card and wall-mapped them in a war room. Two days in, the pattern was visible: roughly eighty per cent of the cards shared one systemic root, a change-management process that hadn't kept pace with a product-mix change on the line. The rest were satellites of that core.

That's when the tools earn their keep. We ran the clustering through 8D discipline — one corrective spine, D2 through D8, with the individual findings mapped as branches — but moved at QRQC speed: same-day containment, daily stand-ups at the wall, red cards turning green. A3 gave the board one page to govern it all. The mistake I see constantly is organisations opening ten separate 8Ds, running ten project managers, and drowning. Ten parallel problem-solving processes signal exactly the fragmentation that caused the findings. One systemic corrective action with rigorous branching signals a system that understands itself.

At SNOP, building a quality organisation for over 900 people at a greenfield plant, the same clustering discipline is how we cut defect costs by 70% and held customer satisfaction at 98%. Not by fixing everything — by fixing the thing everything grew from.

De-escalation is engineered, not negotiated

Escalations have a rhythm. De-escalation has to match it. Zero critical customer escalations within a quarter isn't luck, it's cadence — and leading quality across a 2,000+ workforce, multi-site, taught me the lesson at scale: you cannot talk your way off the ladder. You evidence your way off.

What customers and regulators need is three things, delivered on a clock. First, containment honesty. Say plainly what you've quarantined, what you've shipped that might be affected, and what you don't yet know. Overclaiming containment is the fastest route from VAI to OAI, because auditors test claims, and the gap between your slide and their sampling becomes the next finding. Second, shown system behaviour — not the revised procedure, but the revised procedure with three weeks of execution records behind it. Logbooks, reaction times, escape analyses. The system behaving is the evidence; the document is just the promise. Third, exit criteria defined in the auditor's eyes, not yours. Ask them directly: what would you need to see at the next visit to write a different classification? Then deliver slightly more than that, slightly earlier than promised.

The trade-off is cost. Third-party inspection, expedited validation, dedicated escalation teams — I've signed invoices in the hundreds of thousands of euros for all of it. Cheap, next to a new business hold. A quarter of frozen awards costs more than the entire de-escalation programme.

Key takeaways

  • Classifications are trust verdicts, not arithmetic. VAI versus OAI measures the regulator's confidence in your system's ability to self-correct — your responses shape the classification as much as the findings do.
  • Cluster before you correct. Wall-map every finding back to a single systemic root and run one 8D or A3 spine with branches. Ten parallel problem-solving teams advertise the fragmentation that caused the audit.
  • De-escalate with evidence cadence, not promises. Honest containment disclosure, three to four weeks of shown system behaviour, and exit criteria written by the auditor's eyes — delivered early, every cycle.
  • Price the ladder honestly. Controlled shipping and new business holds cost more than the entire de-escalation programme. Spend on evidence, not on presentations.

The exit meeting is where the classification starts, not where it ends. The auditor who walked out with ten observations has already drafted the next visit's verdict — in pencil. The next thirty days of evidence, or the absence of it, decide whether that pencil gets inked over or erased. The certificate on the wall changes nothing; the logbook on the line changes everything. Caplin Point's sterile facility and Aurobindo's unit now face the same test my suppliers and plants have faced for two decades: show the system correcting itself, or someone else will correct it for you. Quality management, ethical hacking, regulation — one discipline. Understand how systems fail, and build them so they don't.