Mitsubishi Motors: how 23 years of hidden complaints became criminal negligence

Here is an uncomfortable observation from two decades on shop floors: the failure that destroys a company is rarely one its systems cannot detect. Mitsubishi's quality organisation did not lack data. It held complaint records reaching back to 1977 — decades of field evidence naming known defects. What it lacked was a decision process obliged to obey that data. Between the complaint log and the recall decision sat a filter, and the filter's job was brand protection. Anyone who has run an 8D or a QRQC knows the shape. Detection worked. Escalation was captured by the wrong owner.

The situation

By the late 1990s Mitsubishi Motors was Japan's fourth-largest carmaker — proud, financially strained, and shopping for a partner. DaimlerChrysler took its stake in 2000, months before the raid.

Behind the walls, the operating pattern was simpler than a conspiracy. When customers complained about a recurring defect, dealers fixed it quietly during routine service. No recall filed with the ministry, no letter to the owner, nothing in the recall statistics — just a repair note in a file nobody outside was meant to see. Part by part it looked like customer service. Aggregated over 23 years it was a shadow recall programme running parallel to the official one, unaudited and unowned.

How it unfolded

The leak came first. Whistleblowers took the hidden complaint records to the Transport Ministry, and in July 2000 officials raided Mitsubishi's offices. The company admitted concealment stretching back to 1977 and began recalling — eventually well over a million vehicles. Apologies and management reshuffles followed.

Then the case turned criminal. Despite the 2000 cleanup, the truck business carried the same concealment culture, and in 2002 a fatal crash was traced to a clutch defect the company had long known about and left in the field. In 2004 prosecutors arrested senior executives — among them a former president — on professional-negligence charges. DaimlerChrysler refused further support and walked away from the partnership. Sales and the share price collapsed, and the rescue had to come from inside the Mitsubishi group.

Root-cause anatomy

Technically, the clutch defect was not exotic. It was a known failure mode with a documented mechanism — precisely the information a PFMEA exists to hold. The organisation had done the detection work and then quarantined the result. Organisationally, the failure lived in the recall decision gate. That gate is supposed to ask one question: is this defect severe and frequent enough to act? Mitsubishi had substituted another: what does acting cost our reputation? Once reputation is the metric, concealment becomes the rational local choice. Each quiet repair was cheap, fast and career-safe. The liability accumulated off the books.

This is Cost of Poor Quality deferred with a multiplier. The recall cost originally avoided returned as million-plus-vehicle recalls, a dead alliance and criminal prosecution — paid against savings measured per repair.

Concealment never deletes the cost of a defect — it only sets the interest rate.

Where the quality system failed

Start at the PFMEA line. A clutch failure with fatal consequences carries a severity rating that ends all debate; once complaints document that failure mode in the field, occurrence statistics are irrelevant. The required response is containment and a recall decision, not continued repair.

Then the CAPA gate. The quiet repairs were fixes without documented problem-solving, and a repair is containment — D3. Run it without a problem description, a root cause and systemic prevention, and you have a CAPA process in name only. Mitsubishi ran that loop for two decades and called it quality.

Audit independence and KPI design finished the job. Hidden complaint records are exactly what audit access must be built to surface, yet nobody reconciled repair orders against the complaint database — because nobody independent held the authority to pull raw records. And recall counts, treated as a reputation metric, invited gaming. A metric that rewards the absence of recalls gets you the absence of recalls. On paper.

What would have caught it

• A whistleblower channel run as a quality control, not an HR formality — routed outside any hierarchy it can implicate, with guaranteed protection and every entry tracked like a nonconformity.
• Mandatory, auditable defect reporting: every field complaint reconciled to a dated decision log — escalate, repair or dismiss, each with written rationale. Auditors sample the reconciliation itself, not curated summaries.
• A CAPA gate with teeth: no repeat repair on the same defect code without an open 8D or A3 number, and goodwill-repair coding reviewed monthly. A pattern of quiet repairs is a smoke alarm.
• An escalation service level: a known, safety-relevant failure mode forces a documented decision within days — not a quarter, and certainly not 23 years.

My take

This case is the mirror image of everything I have spent two decades building. At SNOP, running quality at a 900-plus-employee greenfield plant, we cut defect cost by 70% and reached zero critical customer escalations within a quarter — not by hiding complaints, but by draining them fast through QRQC and A3. At Witte Automotive the same logic held. Honest escalation inside a quarter is cheap. Forced escalation after a fatality is criminal.

The most expensive sentence I know in manufacturing is let's just fix it quietly for this customer. I have heard softer versions of it in every plant I have ever entered — "let's confirm the failure rate first" — and the control is always the same: make the honest path the documented, auditable, career-safe path. Concealment is a system design outcome. Design against it.

Mitsubishi's engineers detected the defects. The records existed from 1977 onward. What the company built — deliberately, over 23 years — was a decision process immune to its own data. That is the flaw to check for on any floor: not whether you can find defects, but whether anything in your system obliges you to act on what you find. Where that obligation is missing, concealment is not a risk. It is a schedule.