I have sat in a lot of meeting rooms in Shenzhen and Dongguan with a founder on one side of the table, a sales manager and an engineer on the other, and a sample in the middle that nobody is happy with. The founder says the finish is wrong. The engineer picks up the drawing, turns it around, and points at the surface callout. There isn’t one. There is a render, a note that says “premium matte”, and a colour reference from a Pantone book that was never specified as coated or uncoated.
That sample was built exactly to the document it was given. The factory did its job. The buyer did not.
After two decades building things between Europe and Asia, including a consumer hardware venture with Jean-Michel Jarre that started as one product and grew into a line of eight, I have watched this scene repeat with almost no variation. Founders fly in chasing a lower unit price. They leave with a quality problem they created themselves at the RFQ stage, and they call it “Chinese quality”. It is not. It is Western under-specification meeting a manufacturing culture that will build precisely what you asked for, at the price you negotiated, and will not volunteer the twelve decisions you forgot to make.
The gap is in the drawing, not on the factory floor
Here is the uncomfortable mechanic. A Chinese factory quoting a new overseas customer is pricing risk as much as parts. They do not know if you will reorder. They do not know if your volumes are real. They have seen a hundred Kickstarter decks. So they quote to the document, assume the cheapest compliant interpretation of every ambiguity, and keep their margin thin enough to win the job.
Every blank in your spec becomes a decision the factory makes in its own favour. Not maliciously. Economically. If you do not state the resin grade, you get the grade their warehouse already stocks. If you do not state the wall thickness tolerance, you get whatever the tool delivers on a good day. If you do not define the acceptance criteria for cosmetic defects, you have no basis to reject a shipment, and the argument becomes a negotiation about your feelings rather than a conversation about a number.
When we were scaling that hardware line, the difference between the products that went smoothly and the ones that burned six weeks of my life came down almost entirely to how much of the specification was quantified before the RFQ went out. Not after. Before. A spec that says “no visible scratches” is worthless. A spec that says “no scratch longer than 0.3 mm visible at 40 cm under 800 lux on Zone A surfaces, maximum two per unit on Zone B” is enforceable. The first one produces arguments. The second one produces parts.
The same logic applies to packaging, drop testing, cable strain relief, screw torque, adhesive cure time, and the acceptable variance in the click of a button. If it matters to your customer and it is not a number in a document, it is not in your product.
Why the lowest quote is usually the most expensive one
In any china factory sourcing exercise you will get a spread. Send the same package to six suppliers and the range will often be wide enough that founders assume someone is lying. Sometimes they are. More often the spread is telling you who actually read the document.
I have come to treat the highest serious quote as the most informative one. Not automatically the winner, but the one worth a phone call. Factories that quote high on a well-written spec are usually the ones that costed the secondary operations you specified, priced the tolerance you asked for rather than the one they would prefer, and included the inspection labour. Factories that quote very low have almost always stripped something out, and you will find out what during the second production run, not the first. The first run they will hold together with overtime and hand-sorting to win your trust.
Run the numbers over eighteen months, not over the first purchase order. Take the low quote and add the cost of two additional tooling revisions, a four to six week delay, one air freight shipment to rescue a launch window, a 3 to 5 percent field failure rate, and the returns handling that follows. Then compare. In my experience, the cheaper supplier stops being cheaper somewhere around the second reorder, and the gap keeps widening because you have now sunk tooling into a partner you do not trust.
This is also why I have little patience for the logistics-arbitrage conversation founders want to have. If your margin is broken, a smarter 3PL arrangement will not fix it. You gave that margin away at the RFQ stage by buying a price instead of buying a capability. Fix it upstream or accept it.
The first article is more binding than the contract
Vetting chinese manufacturers is not really about the contract. I have had well-drafted agreements that were never referenced once, and handshake arrangements that held for years. What actually governs the relationship is the first article inspection, because that is the moment the factory learns how much you notice.
Treat the first article as a formal event, not a photo exchange on WeChat. Someone competent measures it against every dimension and every cosmetic criterion in the document. You report back in writing, itemised, with numbers. You reject on things that are genuinely out of spec even when they are small, and you explicitly accept the things that are within spec even when you do not love them. That second half matters as much as the first. A buyer who rejects everything is noise. A buyer who rejects precisely is a buyer the factory assigns its better line to.
What you are doing at the first article is setting the standard for the next three years of production. Factories calibrate their effort to your attention. This is not cynicism, it is capacity allocation. Every plant has an A line and a B line, a senior QC and a junior one, and a finite number of engineering hours. Where those go is decided by which customers create consequences.
On vetting more broadly, I put far more weight on what I can see than on what I am told. Ask to walk the floor unannounced on the second visit rather than the first. Ask which of their existing customers is closest to your product category and volume, then ask what their reject rate is on that programme and watch whether the answer is a number or an adjective. Ask who owns the molds for that customer. Ask to meet the engineer, not the sales manager, and see whether the engineer has read your drawing. Fifteen minutes with the person who will actually run your tool tells you more than any audit report.
Tooling, ownership, and the leverage you forget you have
Tooling is where relationships quietly go wrong. If the factory pays for the mold and amortises it into your unit price, you have bought convenience and sold your exit. Moving production later means a new tool, new validation, and a delay you cannot absorb.
Pay for your tooling separately, own it in writing, photograph it with its identification number, and specify the steel grade and expected shot life in the purchase order. This is not adversarial. Good factories are comfortable with it because it signals a buyer who intends to be around for years. The ones who resist are telling you something useful about how they plan to retain you.
The related question, how to manage a chinese supplier relationship over time, comes down to being predictable. Forecast honestly even when the forecast is small. Pay on the agreed date without being chased, every time, because in a market where many overseas buyers stretch terms the ones who do not become the accounts nobody wants to lose. Visit when nothing is wrong. Give the factory some visibility into your roadmap so their engineering team can bring you ideas instead of only quotes. The best cost reductions I have seen did not come from negotiation, they came from a process engineer who understood where the product was going and proposed a change to the part geometry.
When you put a person on the ground
Founders ask me at what point they need someone in China. My honest answer is earlier than they want to hear, and the trigger is not revenue, it is complexity. One simple product, one factory, stable design, you can run it with a good third-party inspection firm and a disciplined travel schedule. The moment you have multiple SKUs, a second supplier, sub-assemblies moving between plants, or an engineering change cycle running faster than your flights, remote management stops working. The cost of a capable local engineer or quality manager is small against one bad container.
Language is part of it, but the real value is presence. A person on the ground finds out on Tuesday that a sub-supplier changed a component, instead of finding out on inspection day. That is worth more than any clause.
None of this is about being tough with suppliers. It is about removing ambiguity, because ambiguity is the single most expensive thing you can put in a purchase order. Write the spec as if the person reading it has never seen your product, will never speak to you, and is optimising for their own cost. Because that is exactly what is happening, and it is entirely reasonable that it should be.
I write from twenty years of building businesses between Europe and Asia. If your company is facing this, start a conversation.