Larry Qi · Company & era

My first job was at Toshiba, in the Hangzhou plant, in purchasing.

The company gave me an excellent start.

Clear processes: the ERP was Oracle, a submarine cable ran straight between the Japanese head office and the Hangzhou plant, and every order and every receiving inspection left a trace in the system — who did it, when. Accurate data: receiving, inventory, payment; the books were the books and the goods were the goods, and it all reconciled at month end. Improvement was pushed daily: defect handling, supplier evaluation, every one with a process, a form and a follow-up.

The team hired fresh graduates; after a year or two, whoever had a better offer left; and every time somebody left, Manager Ishikawa and the two section chiefs quietly pooled their own money to take us all out to dinner — never on the company account.

I learned a great deal there. The fundamentals of purchasing were drilled into me in that factory.

But there are two things I have carried with me ever since.

1. The month-end sheet

The sheet came at the end of every month. In Japanese: 強制払い出す — compulsory issue.

Cartload after cartload came out of the third-party warehouse and was pulled into our own plant. The sheet arrived; we signed. Once signed, that material was our inventory — and whose warehouse the goods sit in decides whose books they land on.

And before that material came in, there was another set of rules plainly in place: VMI, vendor-managed inventory — goods in the third-party warehouse, drawn as used, settled as used. I had a VMI target myself: the external-warehouse ratio, to reach 68% at period end.

We reached it.

The compulsory issues I saw in those months were roughly these:

The largest single one was Intel CPUs: one sheet, 11.18 million dollars, 213,199 pieces, not one short.

On one side, a beautifully written set of rules. On the other, a sheet that came every month regardless. What that system looked like as it actually ran in this plant was formalism.

2. The door opens in Japan

Where the goods on the sheet came from was not ours to decide.

Which panels we bought was decided by the Japanese head office; we executed. Later, going through the file copies of those compulsory-issue documents, I found the remarks column carried the same phrase every time: "OME request."

There was another detail in the supplier list. Those foreign companies were all written down under their Japanese legal entities — SAMSUNG JAPAN, LG.PHILIPS LCD JAPAN, Hynix Semiconductor Japan. Even Tyco, an American company, appeared as "Tyco Electronics AMP K.K."

To come in through this plant's door, you first had to have a legal entity in Japan.

I once broke the purchasing table down by category and did the arithmetic:

Those six together accounted for a third of our purchasing spend.

Add the Japanese share in panels and other categories, and more than half of this plant's purchasing spend went to Toshiba group and Japanese partners.

A plant built in China, with its six most critical tenths of purchasing locked inside the group and Japanese partners. The list was written by somebody else and the threshold stood in another country — that is what a closed purchasing system looks like.

3. What is said inside cannot go up

Outsiders cannot get in: that is a matter of system. What is said inside cannot go up: that is a matter of culture.

In 2015 Toshiba had its crisis. Three presidents resigned together; eight of the sixteen senior executives left.

A third-party committee of outside lawyers and accountants investigated for months and produced a report of more than two hundred and ninety pages. The report does not mince words:

"Toshiba had a corporate culture in which a superior's wishes could not be defied. Without the superior's approval, normal accounting treatment could not be carried out."

"Correcting the practice of padding profits was, in effect, impossible."

The report established that three presidents pushed targets down under the name of "challenges"; that when targets were missed, the president would demand of subordinates why they had not been met, and press them to overstate; that one of them routinely applied pressure to subordinates directly by e-mail or telephone. Senior management knew profits were being inflated and did nothing to stop it.

Of the 151.8 billion yen overstated, the report named three blocks: transport systems 47.7 billion, the television and computer business 68 billion, semiconductors 36 billion. The most heavily padded was television and computers.

The report's final characterisation: "The entire organisation, including senior management, was involved."

And the case began with an internal whistleblower — before that, not one word of what the report describes had been said in a meeting.

It was not that nobody saw it. It was that seeing it did not entitle you to say it. That culture has an exact name: obedience upward, and nothing else.

4. Three faces of one thing

I have described three things. Put them side by side.

The system was form — rules written beautifully, sheets arriving on schedule, targets met. The structure was closed — six tenths of purchasing locked inside the group and Japanese partners, the door opening in Japan. Speech could not go up — the two sentences from the report stood, and stand.

These are not three separate faults sitting next to each other. They mesh:

Because the structure is closed, outside pressure cannot get in. Supplier competition, market shifts — none of it reaches the table where decisions are made.

Because speech cannot go up, inside problems cannot get out. A defect the front line can see never reaches anyone able to fix it.

Both ends plugged, and only form is left in the middle. The system runs, the targets are reported, the meetings are held — everyone "did it right," and the machine had already come loose from reality.

So formalism is not laziness. Inside a closed system it is the most rational way to survive.

5. The bill

The bill for that did not arrive the same day.

In those years the industry outside was changing dynasties: panels moved from Japanese to Korean and Taiwanese hands; PC brands dropped out one by one — Hitachi, Sharp, Sony, NEC, Fujitsu, and finally Toshiba itself.

The chain ran like this —

Because it was closed, outside price movements could not be transmitted in. Panels fell 22% in half a year; the purchasing structure did not move a millimetre.

Because speech could not go up, the problems the front line saw could not be fixed. The television and computer business, the report said, padded 68 billion — the largest of the three blocks.

Because everything ran on form, the company kept making decisions on data that "looked normal" — until the day outsiders saw that what it said and what it did no longer matched.

A company that took a hundred and forty years to build needed less than ten to strip those businesses out.

To close

I left that seat long ago, and the plant has changed owners several times since.

I am not writing this to talk about Toshiba.

I am writing it to say this: the most dangerous moment for a system is not when it breaks. It is when it is still running normally and no longer has anything to do with the truth.

The sheets get signed, the targets get reported, the meetings get held. Everyone did it right, and nobody did anything wrong.

Until one day outsiders see that what it says and what it does are not the same — and by then it is too late.