From prototype to production

The stages, briefly

Most electronics products pass through the same sequence, whatever the team calls it: a file review that establishes what exists; a prototype build that proves the design; NPI that converts the design into a repeatable process; a pilot run that proves the process; test and validation that defines how each unit will be accepted; repeat production; and, when the product ships as a finished unit, box build. The stages are described individually on the project-stage selector; this guide is about what changes between them.

The changes are not gradual. Each stage shifts the balance of cost, risk, and control in a way that catches teams by surprise if they expect production to be “prototype, but more.”

Quantities and lead times

At prototype, quantity is whatever the engineers need: five boards, ten, twenty. Lead time is dominated by the slowest single component and the fabrication turn. Expediting is normal and affordable because the fixed costs are small.

At pilot and production, quantity is set by demand, price breaks, and minimum order quantities. Lead time becomes a plan rather than a wait: components are ordered against a schedule, boards are fabricated to arrive with them, assembly is booked into capacity, and test follows. The critical path is almost always component lead time, which can exceed every other step combined. Expediting still exists but is now a decision with a cost, not a habit.

Practical consequence: state expected annual volume with your first request, even approximately. It changes how the review plans sourcing and tooling, and it is far easier to plan a production relationship from the start than to convert a series of one-off orders into one later.

Cost structure

Prototype cost is dominated by fixed items: fabrication setup, stencil, programming setup, engineering time. Per-unit material and labor are small in comparison, and the price per board is high because those fixed costs are spread over a handful of units.

Production inverts this. Fixed costs are amortized across the run and per-unit cost is dominated by components, then by assembly labor and machine time, then by test. This shifts what is worth optimizing. Panel utilization, which barely matters at ten boards, is a real lever at a thousand. A component that costs a little more but comes on a reel with no minimum-order penalty may be cheaper in production than a bargain part sold in cut tape. A test fixture that is absurd for a prototype pays for itself over a production year.

Sourcing risk

A prototype BOM can be filled from whatever is in stock this week. A production BOM has to be fillable next quarter and next year. The risks that appear at scale:

• Single-source parts — a microcontroller or sensor with no drop-in alternate. Acceptable if known and planned for; dangerous if discovered at the first shortage.
• Lifecycle status — parts marked not-recommended-for-new-design or end-of-life. The prototype worked; the production run cannot buy the part.
• Minimum order quantities — fine at a thousand units, awkward at fifty, where a single reel may exceed a year of demand.
• Allocation — market-wide shortages that put lead times at months for otherwise ordinary parts.
• Counterfeit exposure — the temptation to buy from brokers when authorized channels are empty.

The mitigations are procedural: a BOM review that checks lifecycle and availability line by line; approved alternates recorded before they are needed; long-lead lines ordered ahead of the board schedule with your approval; authorized-channel sourcing as the default, with broker purchases only by explicit decision. See the BOM checklist.

Documentation and control

A prototype is documented in the heads of the people who built it. That is fine for a prototype. Production requires that a facility — possibly one you have never visited — can build the product correctly from documents alone: released fabrication data, a controlled BOM with alternates, an assembly drawing, process notes, an inspection plan, a test procedure, and packaging requirements. NPI is where this set gets completed; the NPI guide describes it in detail.

Control means that changes go through a process. A part swap that an engineer would make on a prototype bench in ten minutes becomes, in production, an engineering change with a revision, an effectivity, and an approval. This is not overhead for its own sake. It is what allows a field failure to be traced to a build, and a second facility to be brought on without relearning the product.

Test

Prototype test is exploratory: the engineer probes, measures, and forms an opinion. Production test is a procedure with pass and fail limits that someone else executes, usually many times a day. The transition is a design task of its own: deciding what “working” means as observable checks, choosing methods, and setting limits with margin for production variation.

Test also changes what the board needs: test points for in-circuit test, a programming header or pogo-pin footprint, an interface the fixture can talk to. These are cheap to add during layout and impossible to add after the revision is locked. If production test is even a possibility, design for it during DVT.

When to lock the design

The design should be locked when it has passed its own validation — functional requirements, environmental and regulatory testing as applicable — and when the DFM and DFA findings from the selected facility have been dispositioned. Locking earlier means absorbing changes informally, which corrupts the documentation. Locking later means paying for tooling and fixtures against a moving target.

Signals that the design is ready to lock: the last prototype spin changed nothing that affects fabrication or assembly; the BOM has approved alternates on the lines that need them; test intent is defined; and the remaining open items are cosmetic or software. Signals that it is not: an unresolved footprint issue, a component without a second source in a critical path, or a regulatory test not yet passed that could force a layout change.

Schedule realities

Production schedules are sequential where prototype schedules were parallel. Components must be on hand before assembly; boards must be fabricated before that; and both depend on a released revision. The steps that can overlap — ordering long-lead parts while the layout finishes, fabricating boards while parts arrive — need decisions and approvals to start early, which is one reason a single accountable planner matters.

Build in checkpoints: revision lock, component arrival, first-article approval, pilot review. Each is a point where the plan can be adjusted before money is committed to the next step. A schedule with no checkpoints is a forecast, not a plan.

Asking for quantity breaks

A quote priced at one quantity tells you little about the next order. Ask for two or three quantity breaks that bracket your realistic demand — for example the pilot quantity, a first production lot, and a plausible annual total — and ask for the non-recurring items to be listed separately from the per-unit price. The separation shows you what is tooling and setup (paid once per revision) and what is truly per unit, which is the number that matters for product cost.

Ask also how the quote treats surplus material. Minimum order quantities and reels mean that components are bought in excess of the placed quantity; whether the surplus is held for your next build, invoiced to you, or absorbed changes the effective price. A quote that is silent on this is not wrong, but it is incomplete, and the question is easier to settle before the order than after.

Finally, state a validity period you can live with. Component prices move; a quote that must hold for six months will be priced defensively. A shorter validity with a clear re-quote path is usually cheaper and more honest.

Choosing a partner model

Three models serve different stages and team sizes, and it is reasonable to use more than one over a product’s life.

• Self-service quick-turn services excel at prototypes: upload, pay, receive boards. They are fast and inexpensive for small quantities but offer limited review, limited sourcing control, and little continuity into production.
• Large traditional EMS providers excel at volume, with deep process engineering and supply-chain capacity. They are optimized for large, stable programs and may not be a fit for teams at pilot or low-volume scale.
• Managed manufacturing partners sit between the two: file review and accountability of a single point of contact, with production placed at qualified facilities matched to the project. This is the model Atherton Forge operates. It suits teams that have outgrown self-service ordering but are not yet a volume account — hardware prototypes, low-volume assemblies, NPI and pilot builds, and repeat production for small and mid-sized OEMs.

Whatever the model, ask the same questions: who is accountable when something goes wrong; how are substitutions approved; where is manufacturing performed and how is that confirmed; and what documentation will exist at the end that lets you move if you need to.

A stage-gate checklist

1. File review complete — package gaps listed and closed; capability items confirmed.
2. Prototype validated — design meets functional requirements; findings recorded.
3. Design locked — DFM and DFA dispositioned; revision released; BOM with alternates.
4. Pilot passed — first article approved; yield understood; documentation reflects the build.
5. Test defined — procedure with limits; fixtures scoped and built if needed.
6. Repeat production released — orders placed to the controlled revision; change process in force.
7. Box build planned (if applicable) — mechanical and harness documentation under the same control.