YIVAHUBDeveloping Hardware in Dongguan: Prototype Cycles, Tool Cost and Small-Batch Economics
2026-11-04 · YIVAHUB Team
Dongguan's real product is not electronics. It is iteration speed: the ability to turn a drawing into a physical part in days rather than months, which is why hardware companies develop here even when they intend to manufacture somewhere else. If you are planning a product rather than buying one, these are the numbers that decide whether your schedule is realistic.
The work is split across specialists, and that split is the feature rather than the flaw. A tool shop designs and machines the mould. An injection house owns the presses and often not the mould. A die-caster does metal. A separate shop does surface treatment. PCBA is another supplier again, and the component markets in Shenzhen are next door for anything you are willing to buy off the shelf. No single factory does all of it, which is exactly why the belt is fast: each step is somebody's core competence rather than somebody's side job. It also means four or five counterparties, and somebody has to hold the schedule across all of them.
Choose the prototyping method by the question you are answering, because the methods cost wildly different amounts. A resin print answers form in one to three days, the cheapest way to hold something in your hand. A CNC part machined from stock answers fit and function in three to seven days, with no tooling cost, and it can be made in the production material so the part behaves like the real thing. Vacuum casting from a silicone mould answers feel and finish for five to eight days and a few hundred renminbi per part, and it will give you twenty to fifty units rather than one. A steel injection tool is the method you use when you have stopped asking questions: fifteen to thirty days for a straightforward part, and real money before a single production unit exists. Die-casting tools run longer, twenty-five to forty-five days. The most common expensive mistake is cutting a tool to answer a question a print or CNC part would have answered for a fraction of the cost.
Tool cost is driven by five things, and only one of them is the size of the part. The number of cavities, because two cavities roughly doubles the machining and the steel. The part size, which sets the mould base and the press tonnage. The steel grade, with a pre-hardened grade suitable for tens of thousands of shots costing far less than a hardened tool built for a million. The number of slides and lifters, because every feature that cannot be moulded in the parting direction adds a mechanism. And the surface finish, since a mirror finish is hand-polished hours rather than a machining setting. A simple single-cavity tool for a small part sits at the bottom of the range; a multi-cavity tool with slides and a polished finish is a multiple of it. Ask for the breakdown rather than a single number: it tells you which of your design decisions costs the money.
Small batch has a cheap range and an expensive range, and knowing where the line sits saves the most money. Injection moulding gets dramatically cheaper per part as volume rises from a hundred to a thousand pieces, then flattens: beyond roughly five thousand the curve is nearly flat, so savings from a larger run stop being worth the inventory. Below the line, a machined or cast part usually beats an injection part once you include the tool. The working rule we use is that under about two hundred units of a simple part, machine it or print it; between two hundred and two thousand, a low-cost tool or vacuum casting is worth pricing; above two thousand, injection almost always wins. The exact crossover depends on the part, and it is worth asking a supplier to price all three so you can see your own line.
A trial run of five hundred to two thousand pieces is normal here, and a hundred to three hundred is achievable with the right concessions. The concessions are almost always the same two: accept a stock colour and accept a standard material grade. A stock colour means no colour-matching run and no minimum pigment order; a standard grade means the material is already in the supplier's inventory rather than ordered in. Both are invisible at trial volume and both can be changed at the production order, which is the point of a trial run.
The hidden cost in small-batch work is changeover, and batching is the lever that controls it. Every colour change, material change and tool change carries a fixed setup cost that does not care how few parts you are running. So the cheapest small batch is one that uses one material and one colour, and the most expensive is four variants of a hundred pieces each. If you need variants for a market test, run them as one production block with the changes sequenced inside a single setup window rather than four separate orders weeks apart, and setup is paid once instead of four times.
The component shortcut is the belt's biggest gift to a small hardware project, and it is easy to miss. Because the Shenzhen component markets are next door, you can build a product out of modules that already exist: a standard cell, a standard connector, a standard charging board, a standard sensor. Every module you specify instead of developing removes a design cycle, a tool, a test round and, crucially, a certification burden, because standard parts often already carry the approvals your market requires. The trade is less differentiation in those areas; for a first version that trade is usually correct.
Certification belongs on the schedule before the tool, not after the first production run. A product with a radio needs type approval for the destination market. A mains-powered product needs safety certification. A product with a lithium cell needs transport certification before it can fly. Each is weeks of testing and a real cost, and each is far cheaper to design for than to retrofit. Buyers who plan the tool first and the certificate second find that a component choice made in month one made the approval route expensive in month six.
Freeze the design before the tool, and expect the freeze to be uncomfortable. A tool is a commitment: after it is cut, every change is a machining operation and a delay rather than an edit. The discipline that works is to answer the form, fit and function questions with prints and machined parts first, get a physical assembly in hand, and only then cut steel. Projects here usually stall for one reason: a design was still moving while the tool was already being cut, and the tool is what forces the argument.
Where this lands with us. We run the development sequence as a schedule with named gates: prototype method chosen by question, tool quotation broken down by cost driver, trial run sized to the cheapest useful volume, certification scoped before the tool is cut. Send the product, the quantity and the destination, and the plan comes back with tooling cost, sampling timeline and certification route separated rather than bundled.