The memo is lean. The memory isn't.

"Overstaffed." That is the word Boeing chose this week – an internal memo, 2,500 positions, roughly 5% of the workforce – while the quality recovery stands on footing you do not jog on. Somewhere a board is already drafting the question: Boeing did 5%, can we do 5% too? Here is an uncomfortable observation from two decades in automotive and aerospace quality: I have never seen anyone cut their way out of a quality crisis. Only defer the invoice. A flat 5% is not an org-chart edit; it is a PFMEA nobody ran on the organisation itself. Severity 9, occurrence near-certain, detection nil.

What actually leaves in a flat 5%

Every plant runs on two org charts. The first lives in SAP and shows who reports to whom. The second exists nowhere and shows who knows what – and the second gates the escapes. The shift leader who knows press 4 drifts two tenths once humidity crosses 70%. The supplier-quality engineer who reads one supplier's certificates with context the certificate does not contain. The inspector who can tell a good weld from a nearly-good one by ear. A flat percentage operates on the first chart and amputates the second, which has no severance formula.

The arithmetic gets worse from there, because cuts are never clean. You cut 5%; a further slice of your best people resign within two quarters, because people with options exercise them first. What remains is survivor silence. Near-misses go unreported. Knowledge gets hoarded, because being the only one who understands a containment plan now feels like job security. That behaviour is rational. That is the problem.

I know what quality capability costs to build, because I once paid retail for it. At SNOP I built a QA/QC department for a 900-person greenfield plant from an empty office – no validated procedures, no failure history, no supplier scar tissue. It took years, and every useful line in it was a scar somebody earned. That is the inverse lesson the memo misses: acquisition takes years, erasure takes a quarter and one signature.

The echo you hear eighteen months out

The escape comes back 18–24 months later. Not because the fix was wrong – the conditions that bred it recur. A tool change, a supplier batch, a parameter someone optimises without knowing why the limit sat where it sat. This time nobody in the room remembers closing it. The new team opens a fresh 8D while the solved A3 sits in an archive nobody searches, or on a laptop IT has wiped. Full tuition, repaid, for a lesson the organisation already owned.

Headcount recovers in a hiring cycle; failure memory recovers in a failure cycle.

Do not expect the audits to catch it. AS9100 and VDA 6.3 sample the system, not the memory. An auditor verifies that containment records exist; no auditor verifies that anyone left in the plant could re-execute them. A competency matrix full of names passes beautifully right up until the names leave.

At Witte Automotive I drove a substantial failure-cost reduction with QRQC and A3, and the honest accounting gives the tools half the credit. The other half was a veteran core who had seen the failure modes before – people who could say this smells like the crimp issue from two winters ago and cut three weeks off an investigation. The quarter I later spent at SNOP without a single critical customer escalation was carried the same way: people who recognised an escape before it became a phone call. Procedures recognise nothing. People do.

The two cost lines never meet in one report, which is why the memo gets approved. A saved €70,000 salary is immediate and auditable. The replacement arrives eighteen months later as a €200,000 sorting, containment and expedited-freight bill on a different budget line. No document connects them. The spreadsheet, unlike the press shop, rounds in your favour.

How to cut without amputating

If the board insists on a number, earn the right to shape it. Having led quality across a 2,000+ workforce over multiple sites, the first artefact I would build before any reduction is a knowledge map: which processes, suppliers and failure modes live in fewer than three heads. Three is minimum redundancy; any failure mode whose containment lives in one head is a no-cut zone, whatever the percentage says. Then triage by criticality, not span of control. Some management layers are genuinely fat – a flat 5% finds them by luck, not method.

Make memory institutional before anyone becomes redundant: searchable A3 and 8D archives, closure interviews asking the leaver what they would check first if the escape returns, handover notes written by whoever owns the scars. Weeks of effort, offset by a single re-learned escape.

A note to the recruiters scanning the aerospace CVs arriving by the hundred: interview for memory, not ownership. Do not ask what they ran; ask for three escapes they closed and what each taught them that the documentation does not capture. Hire the candidate who can describe a machine's signature unprompted. Someone else has already paid their tuition.

An "overstaffed" memo is the only corporate document I know that lowers payroll while raising defect cost, and it will be copied precisely because the two lines never meet in one report. When the question reaches your plant – can we do 5% too? – the honest answer is that the question is priced wrong. Fresh eyes, meanwhile, is what a board calls the replacement of eyes that have seen something. Quality capability took me years to build line by line; one memo can erase the better part of it in ninety days. The plants that recover first will not be the ones that cut deepest. They will be the ones that knew, before the signature, which 5% was holding the failure memory – and cut around it.