A founder I know sent me a link a while back. Same marketplace where he sells, same enclosure, same colourway, same button layout, roughly 40% of his price. He was furious about the theft. I asked him three questions. Who owns the injection moulds? Where do they physically sit? How many factories received the full CAD package during your quoting round?

He owned nothing. The moulds sat on the supplier’s floor. Five factories had the complete files, because that is what his sourcing agent said you do to get competitive quotes.

Nobody stole anything. He ran a nine-month process to transfer his product to a company better capitalised, better connected and better positioned than he was, then signed a contract asking them politely not to use it.

I have built consumer hardware in China. With the Jean-Michel Jarre venture we took one product and grew the line to eight, which means eight tooling programmes, eight BOMs, eight sets of supplier conversations that could have gone wrong. I have also spent years in Hong Kong and on the mainland watching other people’s supply relationships from close range, in banking and in healthcare, where the pattern repeats in different clothes. The conclusion I keep coming back to is unpopular with lawyers and uncomfortable for founders: IP does not usually walk out the door because someone breached a clause. It walks out because of how the commercial relationship was architected, and the architecture was set before anyone signed anything.

I am not telling you to skip the paperwork. Register your trademarks in China first, in Chinese characters as well as Latin script, before you talk to anyone, because the system is first to file and squatters read the same trade show directories you do. Use an NNN agreement drafted for Chinese courts in Chinese, governed by Chinese law, with a liquidated damages number and a Chinese entity as counterparty. Get customs recordals so exports of the copy can be stopped at the port. All of it is worth doing and it is cheap relative to a tooling programme.

But run the enforcement scenario honestly before you rely on it. You are a foreign company. The counterparty is a domestic employer in a specific city with a specific tax contribution and a specific number of jobs. Litigation is slow, procedural, and denominated in a currency your board did not budget for. I have watched companies with far more cash than a hardware startup decide, correctly, that eighteen months of legal spend to chase a factory in a second-tier city was worse economics than walking away. If your annual purchase order is a rounding error on that plant’s book, you are asking a court to defend a relationship the factory already valued at close to zero.

Contracts are a deterrent. They raise the cost of copying you and they give you something to point at. They are not a lock. Treat them as one input into a system, not the system.

The five design choices that give the product away

Look at the failures I have seen and they cluster around the same handful of decisions, all made early, all made for good short term reasons.

The first is quoting behaviour. Founders send complete 3D CAD, full bill of materials, firmware architecture notes and assembly drawings to a list of factories to get comparable pricing. That is five copies of your product distributed to five companies that owe you nothing, in exchange for a spreadsheet. You have paid for the quotes in the only currency that mattered.

The second is tooling. If the factory paid for the moulds, or if you paid for them but the invoice does not describe them as your property with mould numbers listed, or if they sit on the supplier’s floor with no removal clause, you do not have a supplier. You have a landlord. The moment you push on price or quality, the answer is a shrug, because they know the switching cost is the cost of retooling from scratch.

The third is concentration. One partner holding the complete BOM, the firmware and the final assembly step has, by definition, the ability to build and sell your product without you. Not because they are dishonest. Because you handed them a complete, functioning business unit and asked them to run it at a margin they can compare, every month, against the margin of running it for themselves.

The fourth is scale. Volume buys you attention, priority and a reason to behave. If you are a small line item, the calculation changes. I have sat in plants where our order was genuinely modest against what came off those lines, and the only thing that kept us relevant was that we were the account nobody wanted to explain losing.

The fifth is absence. Founders who have never walked the floor, never met the owner, never sat through a dinner, are managing a supplier by email through an agent. In that setup you have no early warning, no informal channel, and no personal cost attached to copying you. Nobody has to look you in the face.

The architecture that actually holds

Start with split sourcing on whatever is genuinely differentiating. The point is not to punish anyone, it is to make sure no single party holds a complete picture. In practice this means the critical module, the custom sensor, the acoustic assembly or the firmware sits with a different partner from final assembly. Firmware in particular should be flashed by you, or by a partner you control, at the last possible step, and it should be locked and read protected. Yes, it adds logistics cost. Compare that cost to a clone at 40% of your price.

Own the tooling, in writing, with mould numbers, and know exactly where each mould is. Photograph them with the number plate visible. Include a written right to remove on notice. Better still, once volume justifies it, move to a second tooling set at a second plant. When I have had that in place, price conversations changed tone entirely, because the supplier understood that transfer was an operational task rather than an existential one.

Release information in stages. Quoting does not require full CAD. It requires enough geometry, tolerances, material spec and volume forecast to price the job, which is often a redacted package or a representative sub-assembly. Full data goes to the shortlisted partner after the NNN is signed with a Chinese entity, not before. Slower, yes. Slower is the price of keeping the design.

Be a customer worth keeping. This is the most underrated defence and the one that costs nothing. Pay on the agreed terms, every time. Give forecasts that are honest rather than flattering, and correct them early. Do not change specification three weeks into a run and expect the schedule to hold. Do not grind the price to the point where the account is barely profitable, because an unprofitable customer is one that a factory has already emotionally written off. A supplier who makes a fair margin, on time, on a growing volume, has an economic reason to protect you that no clause can generate.

Then show up. Western executives file this under soft and skip it, and it is the actual enforcement mechanism. Guanxi is not a mystical concept, it is the accumulated record of whether you behave predictably and whether you treat people as counterparts rather than vendors. Visit the plant. Meet the owner, not the sales manager. Learn who really decides. Eat the dinner and stay for the second hour. The value of that is informational as much as relational: the person who tells you that another customer is asking odd questions about your mould is never going to put it in an email.

The question nobody wants to answer

Here is the part founders resist. If a factory can copy your product and win with it, the product was never the moat.

Physical design is copyable. That is not a moral failure of anyone in Guangdong, it is the nature of injection moulding and reverse engineering. What is not easily copyable is a brand people ask for by name, a distribution position that took years to build, a service and warranty operation, a data or software layer that improves with usage, regulatory certifications, and a product line that moves fast enough that today’s copy is a competitor to last year’s version. When we went from one product to eight, the line itself became a form of defence, because a copy of any single item competed with a catalogue and a brand rather than with a company.

So ask it plainly. If your entire advantage is a shape and a price, a copy is not an accident waiting to happen. It is the correct market outcome, and it will arrive from somewhere whether or not your supplier is the one who ships it.

What to do differently is not complicated, it is just unglamorous. Sign the agreements and register the marks, then assume you will never enforce them. Design the commercial relationship so that copying you is expensive, incomplete and less profitable than serving you. Own the steel. Split the knowledge. Release information in stages. Be the account nobody wants to lose. And get on the plane, because the relationship you have never invested in is the one that will not hold when it matters.

I write from twenty years of building businesses between Europe and Asia. If your company is facing this, start a conversation.