Prototype Before You Tool: Hardware's First Lesson

Article

May 1, 2023

Schematics are finished, the BOM is complete, and someone starts talking about tooling and a volume order to save time. The usual result is a very expensive lesson.

A Mistake That Keeps Repeating

Schematics are finished, the BOM just came together, and someone starts talking about cutting tooling and placing a volume order to save time. The usual outcome: the mounting holes are a millimetre off and the board will not fit the enclosure; the power supply cannot carry the full load; thermal design forces the processor to throttle; the assembly sequence is so awkward that the operator stops. One tooling revision costs more than ten rounds of samples; one scrapped batch can eat ten percent of an early-stage budget.

The common feature of that waste is treating an unknown as if it were known. A drawing that looks right and a board that actually fits are separated by a set of problems that only appear when you hold the part. Tooling and volume production exist to replicate a validated design, not to validate one. Reverse the order and cost multiplies.

What Hardware Answers That Drawings Cannot

Samples answer very concrete questions. Whether the holes align and the board drops into the enclosure — only assembly tells you. Power-supply ripple and how hot the controller runs — only measurement tells you. Whether the component layout is sensible — the placement operator will tell you. Whether the device feels right in the hand — only handling says. Every one of these is an assumption at drawing stage.

More importantly, they interact. What looked like a thermal problem changes the layout, which changes the power routing, and the ripple is wrong again. One more sample round walks that chain of consequences to the end; one fewer lets it all surface at finished-product stage. Hardware problems come in layers, and sampling is the shovel that takes them one at a time.

An Order That Holds Up

After schematic and PCB design, build samples quickly and measure them in hardware. If the measurement says yes, revise and sample again. Iterate until the design converges, then move to a pilot run that clears process issues under near-production conditions. Only then talk about volume. Each step is inexpensive, and each removes a class of risk.

The principle is not new; the difficulty is that execution always feels skippable. The usual excuses are time pressure, a customer waiting, a financing milestone ahead. Anyone who has actually run two rounds of this knows that skipped steps do not disappear — they simply wait somewhere more expensive. Sampling money has the highest return of any line in an early hardware budget, because it buys not being wrong.

Make Sampling Cheap Enough Not to Think About

Our plant treats sampling as routine business. Submit drawings or a sample requirement and a solution and quotation come back within twenty-four hours. Common components are stocked, so there is no waiting on an ad-hoc purchase. The factory is its own, so there is no queue behind a subcontractor. A sample can be produced in as little as five hours, and sampling, small batch and volume production sit in the same plant, so process context does not break between revisions.

Once a sample round is cheap enough to ignore, the way a hardware team works changes: instead of debating in a meeting whether an approach will work, you build one and see. The methodology of hardware startups reduces to something simple — get feedback from the real world as early as possible. The shorter the sampling cycle, the faster that feedback, and the slightly better the odds of surviving. Submit the requirement now and look at the solution and quotation before deciding the next step.

When You Can Skip a Prototyping Round

Not every board needs a physical prototype round. If the design is a minor variant of something already built on the same process — a connector change, a firmware tweak, a different value in the same package — the risk does not justify the wait. The useful test is whether any assumption in the design has never been tested on real hardware. If the answer is none, build for production. If there are two or three, prototype them; four or more usually means the design needs another pass on paper before it earns a build slot.

A second case worth naming: enclosure fit. Software simulations of a board inside a housing are often right, but the failures they miss are mechanical — a connector fouling a boss, a keep-out violated by a screwdriver. When the only open question is fit, a single printed and assembled unit is cheaper than a week of modelling. We can advise, from the gerbers alone, whether a build can skip the prototype stage or whether one round of physical verification will save two later.

The Money Saved on Prototyping Reappears Somewhere Else

Teams that cut the prototyping budget usually spend it a month later in another line item: one mold revision costs ten times a prototype run, a returned batch adds freight and rework, and a missed delivery date carries a channel penalty. None of it is labelled prototyping, but all of it is the price of not having built one.

The real value of a prototype is turning unknowns into knowns. Hole positions, enclosure fit, supply regulation, thermal behaviour and assembly feel all get answers the moment the board is in your hand. No drawing substitutes for picking it up and powering it on. Hardware methodology, reduced to a sentence, is getting reality to give you feedback as fast as possible.

For capital-tight teams prototyping has a second effect: seeing the object makes it easier to cut scope. Many teams delete two supposedly essential features after the first sample returns, and the development time saved dwarfs the prototype cost. View the manufacturer's product details