Nidec announced that an internal investigation had found 844 quality-control irregularities. The financial press read a scandal. I read an inventory. Two decades in automotive and aerospace quality tell me how a number like that gets made, and it does not get made by villains. Nobody behind those 844 entries woke up planning to cheat. They woke up short a gauge, behind takt, waiting on a signature everyone agreed was a formality anyway. So they moved. Multiply that by shifts, by months, by sites, and you get 844 of something nobody in the chain would have called wrongdoing.

The daily checks along the way, meanwhile, found zero. Both numbers were accurate, and the gap between them is the entire story. The problem was never concealment. It was bookkeeping nobody did.

One deviation is a decision. Eight hundred is a habitat

Sociologists call it normalization of deviance; Diane Vaughan documented it after Challenger. On a production line it looks smaller than a space programme. It looks like a first-off inspection done from memory because the fixture is occupied. Like a release signature applied at shift end to a batch that shipped at shift start. Each instance has a local reason – takt pressure, a material shortage, a buyer on the phone – and each one is defensible in the meeting where it gets explained. That defensibility is the engine. Nobody aggregates the excuses, so nobody ever faces 844 of them at once. The investigator is just the first person who did the addition.

I watched this drift build site by site while leading quality across a 2,000-plus workforce. Same customer, same standards, same control plan on paper. One plant still lived a different quality reality than another forty kilometres down the road. Not because the people there were worse. Because each site accumulated its excuses in private. Culture is not what a company writes down. Culture is what a site stops counting.

Why your audit never fired

I am fluent in VDA 6.3, so let me translate: a process audit samples documents and moments, while irregularities live in habits. The auditor sees the torque check performed properly during the one hour the line knows it is watched. The audit trail shows signatures. It cannot show that a signature migrated from "before the test" to "after the truck". And once a deviation has been laundered into a local work instruction – a "temporary" reduction in check frequency now in its third year – the audit will certify it as the process.

An investigation works differently because it gets the film, not the photograph: shift-handover notes, rework loops, chat logs, calibration gaps, who knew what and when. That is why an investigation sees in six weeks what the audit system missed in six years. The difference is not rigour.

It is coverage. The QMS samples. The investigator takes a census.

The boring machinery that counts what you excuse

If you want your number before an investigator hands it to the press, the machinery has to be unglamorous enough to survive contact with production. Four pieces. None of them clever.

  • Count daily what you excuse monthly. One board per line: every skip, waiver, late signature, workaround and sorted-rework event, tallied at shift end. Not to punish anyone – to make the population visible. What gets counted daily cannot normalize quietly. It can only be argued with, which is the point.
  • Run QRQC on minors, inside 24 hours. At Witte Automotive we ran QRQC, A3 and a Q-Wall, and most of the failure-cost reduction came not from fixing the big failures but from forcing small deviations to the wall within a day. The mechanism that waves a trivial defect through today is the same mechanism that waves a serious one through next year. Size is not the variable. Discipline is.
  • Write the rules before the first part ships. Building the QA/QC department of a greenfield plant for 900-plus people, I wrote the deviation-review rules before production started, because habits form in a plant's first ninety days and calcify after. Refusing exceptions the time to become customs is a large part of how that plant reached a 70% defect-cost reduction.
  • Price it honestly. A census costs a supervisor twenty minutes a day and a whiteboard. A QRQC on a minor costs five people an hour – €200, call it. The announcement-day phone calls, with customers demanding sorting, re-inspection and containment, run to seven figures before lunch.

An irregularity is not what a plant hides from you. It is what a plant has stopped noticing about itself.

Key takeaways

  • Count daily what you would otherwise excuse monthly – a deviation census makes the population visible long before an investigation makes it famous.
  • Run QRQC on minor findings within 24 hours; the discipline is the defence, not the defect size.
  • Audit habits, not just documents – layered process audits on night shifts, changeovers and unannounced hours, where the habits actually live.
  • Fix the deviation rules in the first ninety days of any new plant or line; culture is cheapest to write when nothing has been excused yet.

Nidec's 844 is doing the industry a favour, though I doubt it feels that way this quarter. Every plant has a number like it: the full count of everything it has excused itself for since the last time somebody did the addition. The only question worth asking is who finds yours first – you, through the boring census and the 24-hour wall, or an investigation, a customer's audit team, a regulator. Their arithmetic is the expensive kind. Yours costs twenty minutes a day and a whiteboard.