2026-10-19 Judgement Seller's notes

How I Read a Factory, From Having Worked Inside One

Peter Yang's own column. He sells chemicals from China, and he has sat on the buying side of the same trade - the judgement here came from buying, and gets used on the selling side. The view is his. SUNWARD INDUSTRY sets the commercial terms, and a price comes as a written quotation.

What counts as evidence is the difference between the day somebody is watching and the rest of the month. I had the unusual position of seeing both, because before I bought anything I worked inside a plant in China as assistant and translator to the industrial director, who was a foreigner and spoke no Chinese. A plant can perform for the watched day, and it can produce its paperwork for a visitor, down to the production record for the batch, because that record is written down after the fact rather than kept live. What cannot be staged is the record for the weeks when nobody was there, and that is the part that tells you whether the factory behind a supplier is real.

What counts as evidence is the difference between the day somebody is watching and the rest of the month, and I had the unusual position of seeing both. Before I bought anything I worked inside a plant in China — assistant and translator to the industrial director, who was a foreigner and spoke no Chinese, which is exactly why he needed one and exactly why I was in the building. That is the only reason I have anything useful to say on this, and what I have to say is mostly about things you cannot see from a visitor's desk.

How do I know if the factory behind a supplier is real?

By what the plant does in the weeks when nobody is there, rather than by what it shows you on the day you visit. A plant can produce its paperwork for a visitor — the certificate, the tidy stores, the prepared account of how the work is done — and it can produce the production record for the batch, because that record is written down afterwards rather than kept live. The record for the weeks when nobody was there is the part that cannot be staged, and it is the part a visit cannot reach. I never had to ask for it, because I was inside the place where it was written.

What working inside a plant taught me

The job was mostly translation, and the translation was mostly for people who did not speak each other's language - the technical team first, and then the industrial director, who was a foreigner with no Chinese. The effect was that I was in rooms where decisions about production were actually being made. A plant has a version of itself that it presents and a version of itself that it operates, and the distance between those two is the whole subject. From inside, that distance is visible in small ways: which steps of the process are genuinely checked, which ones are written down after the fact, and which ones are performed differently on the day there is a customer in the building.

I am not going to list what I learned there, because a list of it would be exactly the thing I do not believe in. But the centre of it is simple. What a plant says about itself is cheap and rehearsed — anybody can describe their own process well. What counts as evidence is the difference between the day somebody is watching and the rest of the month, and inside the plant I saw both.

From a visitor's desk - a clean floor, tidy stores, a certificate, and a well-prepared account of how the work is done.
From inside - which steps are checked, which are only recorded afterwards, which are done differently because someone is standing there, and what happens in a week when nobody from outside is due.

That second line is not something you can pick up on a visit, and I am not offering it as a way of inspecting anybody. It is the thing I know, and it is the reason I am entitled to have an opinion at all.

What the plant writes down when nobody is looking

The clearest example is the traveller card. A batch of material does not move between machines on somebody's memory. It moves with a card attached to it, and the card records when the step was done, which step it was, who was responsible for it, which machine it ran on, and what the material was. The card is not a summary of the batch written up afterwards. It is filled in at the machine, as the work happens, and it accumulates one line per step.

It is worth being precise about the card itself, because there is a part of it that tells you nothing and a part that tells you something. The form is standard — the same layout in plant after plant, because it is a printed sheet, and a printed sheet can be bought by anybody. Part of what goes on it is written by hand, at the machine, by the person who did the step. So the form proves nothing at all, and the handwriting is the part of it that is individual.

Beside it sits the standard operating procedure for that step and the check items that go with it, so the person at the machine is working from something written down rather than from what they remember doing last time. I watched people work to those documents. What interested me was never whether the paperwork was good. It was whether the card in the cabinet lined up with the procedure next to it, and whether the gaps in the card were the gaps where somebody had stopped signing.

In the larger plants there was more of it, and none of it is about quality certificates. Rules about where tools are put down. Safety marks on the floor and on the machines. A nameplate on each machine, and a note on when it is next due for maintenance. None of that says anything about the product. All of it says something about whether the plant is run, because a place where the next service date is written on the machine is a place where somebody is still there next month.

None of that is a secret of that plant, and it is worth saying so, because it is the part people get wrong. A traveller card, a work instruction, a rule about where the tools go, a nameplate and a service date are not that company's invention. They are the ordinary content of lean production — the set-in-place discipline, the standard procedure for the step, the machine identification and the maintenance schedule — and a plant anywhere in the world that is being run properly has them. The rule is global: the same discipline, taught and audited the same way everywhere, whether the plant is in China or somewhere else. So the useful part is not that a plant has them. It is knowing what their absence means, and that only comes from having seen both.

None of that is a checklist, and I am not offering it as one. It is a description of what the paperwork was on the days I was there. The reason it matters to me is narrow. A visitor's photograph of a tidy floor is worth very little, because a tidy floor can be produced on demand. A record written at the machine cannot be produced on demand, and once you know that both kinds of thing exist you can tell them apart.

The same logic applies to anything I photograph on my own supply lines. I take the picture with a camera that stamps the time into the frame, and the frame carries the surroundings as well as the goods, so the record cannot be put together afterwards by somebody with an interest in it. A photograph of a sample with a clock in it is worth more than a photograph of a sample, because it is a fact about when rather than a picture of something. I have wanted to use video for the same purpose for a long time and I have not, because I am not certain the video carries a timestamp I could rely on later. I would rather use the thing that does.

Why there is no complete list for judging a factory

People ask me for a complete list. My answer is that one does not exist, and the first reason is the one above: most of what matters is not on show. Even people with years of experience have no systematic standard for judging a factory. That is not modesty on my part. It is the state of the thing. I have met a great many people who have traded with factories for years and are still working from impressions and habit.

There is a second reason, and it is more practical. A list is a tool for verifying something you have already decided about. By the time you are going through a list, the decision has been made and the list is there afterwards to confirm it. The work that actually protects you happens earlier than that, when you do not yet know what to ask and everything you have been told is plausible. No list can stand in for that stage, because a list cannot ask the question you did not know to ask.

What I have instead is narrower than people imagine. It produces no score and it catches nobody. Mostly it is an ordering of my attention - what I believe, what I am merely accepting, and what I have not checked and have not asked about. A lot of the damage I have seen was not done by people short of information. It was done by people who had information and no way of telling which parts of it they were meant to doubt.

Where it came from is the second part of my working life. After the plant I spent a period doing purchasing myself, and that is where the method formed - not as a set of questions, but as a sense of which question is worth asking next and what an answer that is too smooth sounds like. It does not transfer. I cannot hand it to a buyer as a procedure, or to my own colleagues as a form to be filled in. It is a way of knowing where to put my attention, and it came from having been on both sides of the table.

Is it worth sending someone to look for me

There are multinational websites now where you can hire a local person to visit on your behalf, and I understand the appeal. It is a reasonable thing to do. It will tell you whether the plant runs, roughly how busy it is, and whether the people there can hold a conversation about the material rather than only about the price.

What it settles is less than people hope. A plant that expects a visitor puts its best behaviour on show for the day, which is exactly what a plant does when a customer is in the building. The visitor cannot see the order book. They cannot tell whether the process they have described is a real part of how the goods are made, or something that was written down once for a questionnaire and has been repeated ever since. Those are the two questions that matter most, and they are the two a day's visit cannot reach.

There is a cost as well, and for a first order it is often larger than the exposure it removes. That is not an argument against doing it. It is an argument for knowing what you are buying with it, which is a narrower set of answers than most people assume they are purchasing.

So if you ask me what I would want to know about a plant before I bought anything from it, the honest answer is that I would want to know what the process is like when nobody is looking, and there is no arrangement that gives me that. What I can offer is narrower than that: which questions are worth asking, why they matter, and what a real answer sounds like next to one that is only rehearsed.

Questions this note answers

How do I know if the factory behind a supplier is real?

What tells you a factory is real is what the plant does on the weeks when nobody is there, not what it shows you on the day you visit. A plant's paperwork can be produced for a visitor, down to the production record for the batch, so what is written down is not the thing to read. I was assistant and translator to the industrial director inside a plant in China, and the distance between the version of itself a plant presents and the version it operates showed up in small ways: which steps are genuinely checked, which are written down after the fact, and which are performed differently when there is a customer in the building. That gap is the whole subject, and it is the part a visit cannot reach.

Is a factory visit worth paying for, or is it just a show?

It is a reasonable thing to do, and it settles some questions: whether the plant runs, roughly how busy it is, and whether the people there can answer questions about the material rather than only about the price. What it does not settle is the part that matters most, because a plant that expects a visitor puts its best behaviour on show for the day, which is exactly what a plant does when a customer is in the building. The visitor cannot see the order book, and cannot tell whether the process they have described is a real part of how the goods are made, or something that was written down once for a questionnaire and has been repeated ever since. For a first order it is usually a larger cost than the exposure it removes.

Can you give me a checklist for checking a factory?

No, and it is not a list I can hand over. A list of things to check is a tool for confirming somebody you have already chosen, and by the time you are going through a list the decision has been made. What protects you happens earlier, when you do not yet know what to ask and everything you have been told is plausible. What I have instead is narrower: an ordering of my attention — what I believe, what I am merely accepting, and what I have not checked — and I can tell you which questions are worth asking and what a real answer sounds like next to a rehearsed one.

What records does a factory keep for each batch?

A traveller card that moves with the material and is filled in at the machine, recording the time of the step, the step itself, who was responsible for it, which machine it ran on, and what the material was. The form itself is standard - the same layout in plant after plant, because it is a printed sheet, and a printed sheet can be bought by anybody - and part of what goes on it is written by hand, at the machine, by the person who did the step. So the form proves nothing on its own, and the handwriting is the part of it that is individual. Beside it sits the standard operating procedure for that step and the check items that go with it, so the person at the machine works from something written down rather than from memory. In the larger plants there is more of it again - rules for where tools are put down, safety marks, a nameplate on each machine, and the date the machine is next due for maintenance - and none of that is about the product. What interests me is that all of it is written at the point of work, because a tidy floor can be produced on demand and a card filled in at the machine cannot.

Are the records you describe normal for a factory in China?

More normal than most people expect, and not a Chinese peculiarity. A traveller card, a work instruction, a rule about where the tools go, a nameplate and a service date are the ordinary content of lean production - the set-in-place discipline, the standard procedure for the step, the machine identification and the maintenance schedule - and a plant anywhere in the world that is being run properly has them. The discipline is global: the same one, taught and audited the same way, whether the plant is in China or somewhere else. So the useful part is not that a plant has them, it is knowing what their absence means, and that comes from having seen both.