A founder messaged me last month asking if I knew a good factory in Shenzhen. No spec sheet, no volume estimate, no target landed cost. Just the word “good”, carrying the full weight of a business plan behind it.
I get a version of that message every few weeks now. Sometimes it’s India instead of China. Sometimes it’s phrased as a question about trade shows: is the Canton Fair worth the airfare, should I go to Electronica, is there a show in Chennai I should be at. The question underneath is always the same one, and it’s the wrong question. People think they have a supplier problem, a discovery problem, a “I don’t know where the factories are” problem. They don’t. The factories are not hidden. Alibaba has hundreds of thousands of them. 1688 has more, at better prices, in Chinese. IndiaMART will give you four hundred injection moulders in Gujarat before lunch.
What people actually have is a shortlist problem. They cannot tell the difference between a factory that will build their product and a factory that will quote their product. Those are completely different companies and they look identical from the outside.
Directories and booths are advertising, not sourcing
I spent years building consumer hardware in China, on the venture with Jean-Michel Jarre where one product eventually became a line of eight. In that time I walked a lot of fairs and I clicked through a lot of listings. Here is what I came to understand, and it took me longer than it should have.
A listing on a B2B platform is a paid advertisement. A booth at a trade show is a paid advertisement with carpet. Neither is evidence of manufacturing capability. They are evidence of a marketing budget and, often, of a trading company that owns no machines at all and will subcontract your order to a workshop you will never see, in a town you will never visit, with a quality system that exists entirely in the form of a man named Mr. Wang who looks at things.
This is not a scam. It is a perfectly rational business model. Trading companies exist because small factories are terrible at sales and foreign buyers are terrible at Chinese, and somebody has to sit in between. The problem is not that intermediaries exist. The problem is that a Western founder, standing in front of a booth with a glossy catalogue, has no way to tell whether he is talking to a factory, a trading company, or a factory that is also a trading company for products it doesn’t make.
The same logic applies in India, with local variation. Plenty of serious contract manufacturers around Pune, Chennai and Noida have almost no digital presence, because their order book comes from relationships and they don’t need you. The ones aggressively optimised for English-language inbound enquiry are, as a rule, optimised for enquiry rather than for production.
So the first discipline is this: treat every channel as a source of candidates, never as a source of answers. A directory gives you names. That is all it gives you. The work starts after.
The first visit is a disqualification exercise
The standard founder trip to Asia goes like this. Fly in, visit six factories in four days, come home with samples, pick the one where the boss was friendly and the price was lowest. I have watched this happen many times and I have done a milder version of it myself.
The flaw is that the trip is being run as discovery. You arrive hoping to find a partner, and because hope is a powerful thing, you find one. Everybody is nice to you. The factory floor is swept. The sample looks good because it was made by the best technician in the building, by hand, over two days, and it has nothing to do with what comes off line 3 at four in the afternoon in month seven.
Flip the purpose. The first visit is not there to find a factory. It is there to eliminate two of the three you already shortlisted. You go in looking for reasons to say no, and you only board the plane once you have three credible candidates and a written specification. Three, because with two you have no leverage and with six you have no attention. Written, because a spec that lives in your head is a spec that will be interpreted by someone else’s.
Showing up with neither candidates nor spec is how buyers end up with the number that haunts them: the quote they will never actually pay. You describe the product verbally, they quote a price against the cheapest possible interpretation of everything you left vague, you anchor on it emotionally, and then over the next four months the real cost climbs back by twenty, thirty, forty percent through tooling revisions, material “upgrades”, packaging, testing and the famous engineering change fee. Nobody lied to you. You just asked for a price on a product that didn’t exist yet.
What to look at that almost nobody looks at
Most visit checklists circulating online are about certificates and clean floors. Useful, low signal. ISO 9001 tells you a consultant was paid. Here is what I actually want to know, and most of it you can get by asking directly and watching how the answer is handled.
Who else do they produce for, and at what volume. Not the logo wall in the meeting room, which is aspirational at best. I want the shape of their book: three customers or thirty, domestic or export, consumer or industrial. A factory whose entire output goes to two Chinese domestic brands has processes, tolerances and a cost structure tuned for those brands, and your 5,000-unit export order with CE marking and retail packaging is an exception they will handle badly.
What percentage of their line your order represents. This is the single number I care about most and it is almost never asked. If your annual volume is 2% of their capacity, you are a nuisance order. You will be scheduled last, you will be bumped when the big customer accelerates, and when your spec changes you will be told it is impossible. If your volume is 40% of their line, you have a different problem: you own their survival, which feels like power until a quality issue forces you to leave and you discover you cannot. Somewhere between roughly 10 and 25 percent is the zone where you matter enough to get attention and not so much that the relationship becomes hostage-taking. Ask for their annual output in units and do the arithmetic in the room. How they react to the question tells you as much as the number.
Who in the room actually decides. In most Chinese and Indian factories of any size, the person who speaks English is not the person who controls the production schedule. The sales manager is a translation layer with a commission. I want to meet the plant manager or the owner, and I want to see what happens when I ask a technical question that the sales manager cannot answer. Does he turn around and ask the engineer, or does he invent something? Is the engineer in the room at all? If after two meetings you have still never spoken to anyone who touches the machines, you are negotiating with a façade.
What happens in month four when your spec changes. It will change. In the hardware line I worked on, the product evolved continuously, and every evolution landed on the factory as a request they had not priced. So I ask it hypothetically and in detail: if I change this housing dimension by 2mm after tooling, what is the cost and the lead time. A good partner answers in specifics, something like “that’s a new slide in the mould, roughly two weeks, I’d need to see the drawing.” A bad partner says “no problem, we are flexible.” Flexibility unpriced today is a negotiation you will lose later, from a weak position, with inventory committed.
So the trip is worth it. Just later than you think.
I am not arguing against going. Some things only transmit in person. You cannot smell solvent on a video call, you cannot see that the warehouse is stacked with finished goods nobody is shipping, you cannot watch the owner’s face when you ask about his largest customer. Fifteen years in Hong Kong and mainland China taught me that the relationship layer in Asia is not ceremonial, it is the actual mechanism through which problems get fixed at 11pm on a Saturday. That layer does not form over email.
But the sequence matters. Specification first, written, with tolerances, materials, test criteria and packaging. Then a long list from directories, shows, and far better, from other founders in your category who have already been burned and will tell you who burned them. Then remote qualification: video walkthroughs of the actual line, not the showroom, a request for their output figures, a paid sample order with a deliberately awkward requirement in it to see how they handle friction. Then and only then, three names and a flight.
Do it in that order and the trip costs you a week and buys you a partner. Do it in the other order and the trip costs you a week, a tooling deposit, and about nine months before you realise the price you agreed was never a real price.
I write from twenty years of building businesses between Europe and Asia. If your company is facing this, start a conversation.