What happened?
In 2021, Mitsubishi Electric disclosed quality problems including the use of materials different from those registered for certification by Underwriters Laboratories (UL) at its Kani Factory and failures to perform some customer-agreed tests for railcar air-conditioning units at its Nagasaki Works while still preparing inspection reports.
The company established an investigation committee chaired by an outside lawyer in July 2021 and began a company-wide review. This article uses only the committee’s 291-page first report, published on October 1, 2021.
That report documents not only specific conduct at Kani and Nagasaki, but also the practices that allowed it to continue, why earlier quality reviews did not surface it, and why employees found it difficult to raise problems. Later second, third and fourth/final reports are outside the scope of this article.
Background: Mitsubishi Electric (Wikipedia)
10 passages that stood out
At Nagasaki Works, a program using development-test data and random numbers to generate inspection reports was handed from one test-team leader to the next. Some reports marked waterproofing as passed without the required test, while measuring instruments were adjusted before customer-witnessed testing. These ten passages follow the practical routines, supplier interactions and employee accounts recorded in the investigation.
The source report is in Japanese. This English article is From Reports’ translation and summary, not an official English translation. Page references use the printed pages of Mitsubishi Electric’s published report.
For the UL inspector, suppliers were given false drawings — and even the certified material to show
At the Kani Factory, products were being made with material different from that registered for UL certification. To prevent the mismatch from being discovered during UL’s periodic follow-up inspections, the engineering section asked suppliers to cooperate.
Suppliers were given false drawings stating that the UL-certified material was being used, together with the material that was actually registered with UL.
They were asked to show the drawings and material to the UL inspector and explain that the registered material was in use.One supplier initially refused to cooperate with the false explanation, but the Kani side asked again and ultimately obtained agreement. In 2016, that supplier said it wanted to stop giving false explanations. The Kani side asked it to continue cooperating while efforts were made to correct the mismatch promptly, and again obtained its agreement.
Around 1990, a program began generating inspection reports with random numbers — and was handed down from leader to leader
At Nagasaki Works, mass-produced railcar air-conditioning units were not tested under the conditions specified by Japanese Industrial Standards (JIS) for cooling capacity. Instead, data from development testing was reused to prepare inspection reports.
Orders could contain 20 to 30 units, and preparing reports with many values — voltage, current, temperature, humidity and airflow — was burdensome.
Around 1990, one tester created a program that took development-test data, varied it within tolerances using random numbers, and automatically generated inspection reports for production units.
Use of the program was then handed down from one test-team leader to the next. After Windows was introduced, it was moved to a shared folder so team leaders could use it from their own PCs. According to the report, its use stopped on June 14, 2021, when the program was discovered.
Former design-department employee on production testing
From Reports translation of the Japanese original (not an official translation)
During development testing, we performed many tests to work through risks, beyond the formal tests required by JIS. Everyone focused on development testing. At the mass-production stage, we assumed we were simply making the same product as the one developed.A new leader was shocked by the program, but was told that this was how it had always been done
One former test-team leader told the committee that the program first came to their attention when it was handed over by their predecessor.
They were shocked that the system was not even converting actual measurements: it used figures from earlier development-performance testing to create new inspection reports. They asked whether the practice was improper.
According to this testimony, the predecessor’s response was essentially that this was how they had always done it and they should continue in the same way.
They then inherited use of the program.95% was treated as a pass, and customer-witnessed tests were adjusted to display 100% or more
For cooling-capacity tests during development-performance testing, JIS E 6602 required cooling capacity to meet or exceed the rated capacity — 100% or more. At Nagasaki Works, however, results of at least 95% but below 100% were also treated as passing.
When preparing the inspection report, employees worked backward from the desired result, adjusting temperature, humidity and airflow figures so the report would show cooling capacity at or above 100%.
Customers sometimes attended the performance test. If a test conducted beforehand fell below rated capacity, employees used an adjustment function built into the measuring equipment so the display would show values corresponding to 100% or more, rather than the actual temperature, before conducting the witnessed test.
The report also assessed that these products had no quality or performance problem because they could meet rated capacity under the most favorable temperature and humidity conditions allowed by JIS. It nevertheless noted that the test should have been repeated under those conditions.
The practice stopped for new products but continued for repeat orders, because fixing them could expose earlier deliveries
Former members of the Railcar HVAC Systems Department said that, around 2010, the department head learned that products below rated cooling capacity were being passed and strongly ordered the practice to stop. From around 2011, newly developed products no longer used the 95%-pass treatment or the related report and temperature-display manipulation.
But the practice continued for repeat-production orders of products developed earlier.
The report gives a specific reason: if a design change were made because cooling capacity was below rating, customers might discover that units already delivered had also failed to meet the stated rating.
The practice continued for repeat orders even after a 2019 rule revision explicitly required measured values to be entered in inspection reports.The waterproofing test was not done, yet the report said “Pass”; some models were tested because of customer audits
JIS E 6602 specified a waterproofing test in which a completed railcar air-conditioning unit was sprayed for at least ten minutes with water equivalent to rainfall of at least 200 mm per hour.
Nagasaki Works performed a separate watertightness check on a base component, but did not perform this waterproofing test on the completed unit.
When customers required an inspection report, the waterproofing item was nevertheless recorded as “Pass.”
One former quality-control employee said that the test itself took only about ten minutes, but dealing with the unit after it had been soaked required time and work.
Some products were, however, tested exactly as specified. An employee explained that this was because the customer actively conducted factory audits. The report states that watertightness checks and checks of the sealing gaskets confirmed no performance or quality problem in the affected products.
A reading below 100 MΩ was still written as “100 MΩ” to avoid updating the previous report
For insulation resistance, the internal specification required at least 20 MΩ for the main circuit and 5 MΩ for the control circuit. There was also a rule that readings of 100 MΩ or more could simply be recorded as “100 MΩ.”
At Nagasaki Works, however, even actual readings below 100 MΩ were recorded as “100 MΩ” as long as they still met the applicable minimum specification.
Inspection reports were prepared by overwriting the previous report, and the meter often pegged at its upper limit, so earlier reports frequently already said “100 MΩ.” If the new measurement still met the specification, employees avoided the work of changing the figure and issued the report unchanged.
The report also states that the actual resistance values in these cases met the specification and that no quality or performance problem was confirmed.
Dimensions were not measured; design values were simply scattered within tolerance
JIS did not require dimensional inspection for every product, but some customer contracts did. Even in those cases, Nagasaki Works sometimes did not perform the agreed inspection.
Employees explained that the most important mounting dimensions were tightly controlled with jigs during assembly, so they believed the product dimensions were adequately assured.
The customer inspection report still needed numbers.
Quality Control therefore took design values and varied them arbitrarily within the tolerance range, creating reports that made it appear the dimensions had actually been measured.
A former employee said that dimensional inspections were already being omitted when they began quality-control work in 1985.The report states that accurate component machining and assembly jigs ensured the correct dimensions and that no performance or quality problem was confirmed in the affected products.
The self-inspection sheet marked everything “no risk” — while a system could lower temperature data by 0.5°C
In fiscal 2016, Mitsubishi Electric carried out an internal quality-risk review. On the sheet submitted by Nagasaki Works, cooling capacity, cooling power consumption, mass and dimensions, operating noise and operating vibration for the railcar air-conditioning units all received a final rating of “○ (no risk).”
For cooling capacity, the sheet went further: because measured data and conditions were recorded automatically, it said there was no room for improper manipulation.
In reality, the system could record each one-second temperature reading exported to Excel 0.5°C lower.
The committee said the statement that there was no room for manipulation was contrary to fact.Because every item was rated no risk, no special follow-up or corrective action was taken after the review.
“You lose if you speak up”: proposing an improvement meant getting the extra work yourself
During interviews, employees at Nagasaki Works told the committee that “nothing changes even if you tell management” and that raising a problem meant eventually having to deal with it yourself.
One employee described the workplace with the Japanese phrase “itta mon make” — essentially, “you lose if you speak up.”
According to the employee, the person who proposed an improvement would become responsible for coordinating it, without having the rest of their workload reduced. The result was simply more work. Problems were therefore discussed not in formal settings but only over drinks or in casual conversation.
The committee said this testimony starkly illustrated a failure of middle management to perform its intended role.
Nagasaki Works employee describing managers’ response to suggestions
From Reports translation of the Japanese original (not an official translation)
My supervisor was too busy to listen to my suggestions, so I stopped making them. The department and section managers were so busy that they did not even read emails or documents.What the report says about the background
The head-office quality-assurance promotion department coordinated company-wide quality-assurance activities, but did not have authority to stop product shipments or exercise personnel authority over business divisions (printed page 49, footnote 58).
Every November, quality month included quality-assurance e-learning for all employees. Before fiscal 2016, however, education focused on reducing quality-related costs and improving quality; it placed little emphasis on quality misconduct (printed page 58).
Corrections to this article
Note 04 adds the report’s assessment that rated cooling capacity could be met under permitted test conditions and its qualification that the test should have been repeated (p.202). Note 05 identifies the department head’s stop order as an account from former staff (p.201). Note 06 replaces the exclusive customer-audit wording with the report’s account of some models (p.195). Note 08 adds p.198 for the contract and jig explanations and attributes the account of omitted inspections in 1985 to a former employee.