The first time I sat in a factory meeting room in Shenzhen with a product that existed only as a rendering and a wish, the founder next to me spent forty minutes of a ninety-minute meeting on the NDA. He wanted a clause about employees taking photographs. He wanted to know what happened if a line worker memorised the enclosure geometry. The sales manager nodded politely, signed everything, and then asked the only question that mattered: what is your annual volume?
The answer was 8,000 units in year one, maybe 20,000 in year two if retail worked out. At that point the meeting was effectively over, and not for the reason the founder feared. Nobody in that building was going to clone his product. They were going to accept his order, put it behind three larger ones, and let his tooling sit idle in a corner for five weeks while a customer with ten times the volume took the machine time.
That is the pattern I have watched repeatedly, from consumer audio hardware I helped build in China with the Jean-Michel Jarre venture (one product that eventually became a line of eight) through to conversations I still have with founders every few months. The obsession is IP. The killer is operational.
Why nobody is cloning your 10,000-unit accessory
Cloning is a business decision, and it has a spreadsheet behind it like everything else. To copy your product, somebody has to reverse-engineer the mechanics, cut new tooling, source the BOM, absorb the yield learning curve, and then find a market. Tooling for a moderately complex injection-moulded consumer device with a few slides and inserts runs into tens of thousands of dollars. Add PCB layout work, certification if they want to sell in regulated markets, and packaging. Then ask the crucial question: sell to whom?
Your value is not the geometry of your enclosure. It is your distribution, your brand, your firmware, your community, your retail relationships. If you have 8,000 units of first-year demand and no shelf space, a copy of your product is a copy of a problem. The people with the capability to clone at scale are chasing categories where demand already exists in the millions: charging accessories, earbuds, phone cases, e-bike controllers. Not your niche gaming peripheral with a Kickstarter backer list.
Where copying does happen, and it does happen, it follows demand. If your product breaks out and sells 300,000 units, you will see variants on Amazon and AliExpress within a few months. That is a success problem, and by then you should have brand, firmware, a second-generation product in tooling, and channel relationships that a clone cannot replicate. I would rather have that problem than the alternative, which is that nobody wanted your product enough to copy it.
The exception worth naming: if your differentiation is a manufacturing process, a chemistry, or a piece of mechanical engineering that took years and real money to develop, the calculus changes and you should be selective about who sees what. In climate tech, which is where I spend my days now, that is genuinely the case. Process know-how is the asset. But most consumer hardware founders are not in that position, and treating a Bluetooth accessory like a semiconductor process is how you end up with a signed NDA and a product that is nine months late.
What actually kills first-timers
Four things, in my experience, and none of them involve theft.
The first is an incomplete specification. An ODM quote is a response to what you wrote down. Everything you did not write down becomes their choice, and their choice will optimise for their cost, not your quality. If you did not specify the grade of ABS, you will get the cheapest that passes. If you did not specify the switch manufacturer and part number, you will get a functionally equivalent switch with a different feel and a different failure curve. If you did not specify surface finish with a reference, you will get whatever came off the tool. I have seen a product where the difference between the pre-production sample and the mass production run came down to a single unspecified adhesive, and it took a full container of returns to find it.
The second is tooling ownership. Founders sign contracts where the factory “provides” tooling as part of the unit price. That feels like a saving of thirty or forty thousand dollars up front. What you have actually done is hand somebody a veto over your ability to leave. When the price goes up 12% at the next quote cycle, or when quality slips and you want to move, you discover the moulds are not yours, or they are nominally yours but physically in their building and they have an invoice dispute conveniently outstanding. Pay for your tooling. Get an invoice that lists each tool. Photograph them with their asset tags. Write into the contract that they are released on demand at your cost, and know that enforcement is practical rather than legal: the real protection is a relationship where they want your next order more than they want your moulds.
The third is the absence of a golden sample and a written test protocol. A golden sample is a physically signed and sealed unit, agreed by both parties, that defines what “correct” means. Without one, every quality dispute becomes an argument about intentions. With one, it becomes an argument about a measurable deviation, which is a much shorter argument. The test protocol matters just as much: what percentage of units get tested, on what jigs, against what pass criteria, and what happens to the failures. If you do not define this, “we tested it” means somebody plugged in one unit from the pallet.
The fourth is order priority, and this is the one that no contract fixes. On the factory floor, you are competing for line time, engineering attention, and the good operators. If you are 2% of their revenue and somebody who is 20% needs the line, you will lose, quietly, and you will be told about a component shortage. The only defences are structural: pick a factory where your order is meaningful (being a small customer at a mid-sized supplier beats being invisible at a giant), get a named engineer rather than only a sales contact, and have somebody physically present during the first production run. Not a video call. A person, in the building, for the days that matter.
The real decision: what you own and what you rent
Reframe the question. It is not “do I trust them or go it alone”. It is which layers of the stack you own outright and which you rent from a partner. There are five that matter.
Tooling you own, always, on your invoice. The bill of materials you own, meaning you know every part number and every supplier, not just a lump-sum unit price. The test protocol you own, because it defines your quality standard. Firmware and any keys you own, and you keep production signing outside the factory if the product has connectivity. And second source you own, at least on paper, which means you have qualified a plausible alternative for the two or three components that could stop your line and you know what switching would cost in time and money.
What you rent is everything the partner is genuinely better at: process engineering, line labour, procurement scale on commodity components, compliance testing logistics, and access to a supply base you could never assemble from a laptop in Berlin or Milan. That is a lot of value, and a good ODM delivers it at a cost you cannot match alone.
An ODM quote hides three things, reliably. It hides tooling amortisation baked into the unit price, which means you are paying for moulds you do not own, sometimes twice over. It hides the volume assumption: a quote at 10,000 units assumes 10,000 units and is silent about what happens at 3,000, which is what you will actually order in the first run. And it hides scope, particularly around certification, packaging design, tooling revisions, and engineering change orders. Ask for the quote broken into tooling, unit cost at three volume tiers, NRE, and a per-change engineering rate. If they will not break it out, you have learned something useful.
Questions to ask while you are standing on the floor
Walk the line and ask who else they build for in your category, then ask to see a unit. Ask what their largest customer represents as a share of output. Ask who signs off a first article inspection and to see the last one they issued. Ask what happens when a component goes end-of-life mid-production, and listen for whether they have a process or an anecdote. Ask to meet the engineer who will actually own your project, and if the answer is that sales will coordinate, understand what that means. Ask where tooling is stored and whether you can see somebody else’s moulds with asset tags on them, because a factory that manages other people’s tooling properly will manage yours.
Going it alone (your own contract manufacturer relationships, your own component sourcing, your own mechanical and electrical engineering) is genuinely right in two situations. One, your differentiation sits in the manufacturing process itself, so handing the process to a partner means handing over the company. Two, you already have the operational muscle: somebody on your team has run a production line, speaks the language, and will spend real time in Guangdong. Without one of those, going it alone usually means paying full price to learn what an ODM already knows, eighteen months late.
The founder from that meeting shipped, eventually. Late, over budget, with a plastic finish nobody was happy about. Not one unit was cloned. His problem was never trust. It was a specification thin enough that somebody else got to make the decisions, and a purchase order small enough that nobody on that floor was in a hurry.
Write the spec. Own the tools. Sign the golden sample. Be in the building.
I write from twenty years of building businesses between Europe and Asia. If your company is facing this, start a conversation.