I worked in laboratories in China before I did any business here. Years of it, in Harbin and then in Jiangsu, ordering reagents from domestic catalogues because that was what the budget allowed. It taught me what no reading about supplier assessment could: the name on a bottle and the place its contents were made are two separate facts, and establishing the second one means asking a question nobody expects from a customer.
How do you assess a Chinese pharmaceutical supplier?
Start by establishing whether you are speaking to the manufacturer, an authorised agent or a trading company, because that decides who can answer a technical question at all. Then judge the answers rather than the documents: a supplier whose staff explain their own methods and name their own difficulties is telling you more than a clean certificate does.
That habit is most of what I carried across when I began buying active ingredients and excipients for buyers in Yemen, the wider Middle East and the United Kingdom. People ask me fairly often how to tell whether a Chinese supplier is any good, usually with a quotation in front of them they cannot read. The first honest answer is that it is not a question about China.
Chinese manufacturer is not a category
This country makes a very large share of the world's pharmaceutical raw materials, at every level of care that exists. There are plants where the quality unit has real authority and will hold a batch against the objections of the sales side, and plants where it signs what it is handed. Both send the same style of quotation and describe themselves in the same vocabulary. A judgement pitched at the level of a country discards the only distinction that decides anything, the one between a factory and the factory next to it.
The first question is who I am actually speaking to
An enquiry normally reaches me as a price, a specification sheet and a company name, and the early work is finding out which of three things that name belongs to. It may be the manufacturer, an agent authorised to sell one manufacturer's output, or a trading company buying this month from whoever is cheapest, with a long relationship with the plant or none at all.
None of the three is disgraceful, and I say that as one of them. NJMC owns no plant: we source, we trade and we consult. What the distinction decides is not honesty, it is who can answer a technical question. If material reaches me through two intermediaries, my question about an unexpected peak in a chromatogram travels through two people who cannot evaluate it, and the answer comes back simplified twice.
So I ask for the manufacturing site by name, early, before either side has invested anything in the deal. A supplier who will not say where the material is made has told me what I needed to know.
Questions that separate the two
- Which site makes this material, and is it the site named on the certificate of analysis?
- Who at that site takes a technical question, and how long does a reply take?
- If I order twice in a year, is it the same site both times? An intermediary may change source without mentioning it.
- What analytical method stands behind this specification line? A specification without its method is a number without a meaning.
What a scientist hears in an answer
Here is the part I could not have learned as a buyer. When you have spent years generating data with your own hands, you develop an ear for the difference between an answer from the person who did the work and one from someone repeating what they were told.
The person who ran the assay leads with the awkward part. They tell you recovery was poor at the low end of the range, or that the method struggles to resolve two related impurities. They volunteer the limitation because in their mind it is part of the result. The person who did not run the assay gives you the figure and an assurance that it is fine.
My doctoral work was on how Helicobacter pylori glucosylates cholesterol to protect its cell wall, and on what that does to its resistance to antibiotics. It is the kind of question where a real finding and an artefact look identical for a long stretch. I learned to distrust a clean result years before I had reason to distrust a supplier. Data with no roughness anywhere in it has usually been tidied by somebody, and the tidying is what I want to see.
The language boundary is mine to manage
I move between Chinese, English and Arabic inside a single week, and specifications lose things crossing between them. What survives translation is the nouns. What falls out is the constraints: the temperature a method was run at, the qualifier attached to a stated limit, the difference between a substance looked for and not found and one that was never in the test at all.
Agreement is also the easiest thing to produce in a second language and the hardest to confirm. A yes can mean the words were understood rather than the obligation accepted, and both sides leave the call satisfied for different reasons. The remedy is writing: put the requirement into the purchase agreement in terms that survive being read alone, ask the other side to restate it in their own words, and treat any gap between the two versions as the true state of the agreement.
Being here means I can go and look
The practical advantage of living in Nanjing is not insight into a national character. It is that a factory is a train ride away and a visit needs no month of arrangement.
A hosted tour shows me a clean corridor and a wall of framed certificates. What I want is twenty minutes in the quality control laboratory, reading whether it is a laboratory in use. Are the instruments running? Does the logbook carry handwriting from last week? When I ask an analyst about their own method, do they answer or look at the manager first? A working laboratory looks worked in, and anyone who has spent years inside one sees it from the doorway.
I also ask what happened the last time a batch failed its own specification. Every manufacturer that tests seriously has one. The answer tells me where authority sits: a plant whose quality unit can stop a shipment remembers it in detail, and a plant whose quality unit cannot has nothing to remember.
Where the documents come in
Documents arrive before any of that, and for most buyers they are the only thing that ever arrives. PharmaTrust, the software we built at NJMC, reads a certificate of analysis alongside the specification it claims to meet and reports where the two align and where the specification asks for something the document does not carry. It analyses documents rather than medicines, and it does not replace a quality team or a laboratory. It is silent on everything above, because no document states who signed on whose behalf or what the laboratory looked like. What it does is surface the questions still open while there is time to ask them.
What proximity does not fix
Being here lets me check. It does not make me right. I have been wrong about suppliers in both directions, and the more expensive error was not the plant that turned out to be careless. It was the plant I declined after one bad conversation, which someone else bought from for years without difficulty.
What I hold onto is narrow enough to be honest about. The country is not the variable. The care taken inside one building is, and so is the amount of checking the buyer does, and in the markets I sell into the second is usually the only one within reach. A buyer in Aden cannot audit a plant in Hebei. They can insist on knowing which plant it is, which is the question I started asking when I was buying antibodies on a student budget with no standing to ask anything.