The company that built your five prototypes is often not the company that should build your five thousand units — and the move between them is where budgets quietly disappear. This guide is about recognising the moment, understanding what re-qualification actually costs, and making the change less painful than it usually is.
Eight things that behave differently at 1,000 units than at 5 — most of them not on any quotation.
| What changes | Prototype / pilot run | Production run |
|---|---|---|
| Assembly method | Often partly or wholly hand-placed, with a technician looking at every board. | Machine-placed from a programmed line. No one looks at any individual board. |
| Stencil and tooling | One stencil, made to get you a sample quickly. | Stencil, panel layout and fixtures made to run repeatedly to the same standard. |
| Component source | Whatever is in stock in small quantity — sometimes bought through brokers. | Authorised distribution, one lot per build, traceable to a date code. |
| Programming | Done at a bench, one unit at a time; sometimes not done at all. | A step in the line, with a check that it succeeded before the board moves on. |
| Test method | “Does it power up?” | Functional test against a defined pass/fail, with inspection data you can keep. |
| Documentation | Gerber files and an email thread. | BOM with manufacturer part numbers, assembly drawing, test specification, first-article report. |
| Where the cost sits | The per-unit price is high and nobody is arguing about it. | The per-unit price is the entire conversation. |
| Who fixes a defect | The engineer who built it, from memory. | A documented corrective action — because the same defect will otherwise repeat on every board. |
Answer these honestly before you start interviewing suppliers. Most switches are caused by one of them, not by price.
In the prototype phase, an engineering change order is an email and a rework. Once production tooling exists, an ECO means a new stencil, a new panel layout and possibly a new fixture. A supplier you can stay with is one where the layout has stopped moving — if you are still changing the board every month, no supplier change will fix that.
Prototype shops are organised to build tens. If your forecast is hundreds to thousands and climbing, you are asking a building that was never laid out for it to scale — and the symptoms are the ones you have probably already noticed: lead times slipping, prices revised upward, questions coming back to you that should have been answered internally.
Gerber files are not a production package. You also need the BOM with manufacturer part numbers, an assembly drawing, a test specification, and a clear written statement of what “good” means for your product. If any of that exists only in the prototype engineer's head, you will pay to have it written again at the new supplier — and you will discover the gaps at the worst possible moment.
At low volume nobody asks. At volume your own customer starts asking for inspection data, a first-article report, traceability and a corrective-action record. If your current supplier cannot produce them, the switch is not a decision you are making — it is a deadline you are approaching. Which certifications your product needs →
The quoted part is the smaller half.
The visible costs are straightforward, and you should expect to pay them at any new supplier: new stencils, a new panel layout, test fixtures if the volume justifies them, a first-article inspection, and the engineering time to review all of it. These usually come back as a one-time NRE line, and they are negotiable in the ordinary sense — quantity, timing, whether a fixture is worth building yet.
The larger cost is not on the quotation at all. It is the defect that a prototype build never had to expose. A prototype is built by people who look at every board; a production line is not. So the failure modes that appear at volume are the ones that were always present but were never given the chance to matter: a part that tombstones on a panel with a different thermal mass, a solder joint that passes inspection and fails after fifty thermal cycles, a connector that is 0.1 mm out on a hand-placed board and consistently out on a machine-placed one.
Both lists are real. The second one is the one that saves money.
Five things, in order, that decide whether the new supplier's first build matches the old one's last.
Hongrong builds from 1 piece with no minimum order quantity, and the same site in Bao'an builds up to 10,000 units on the lead-time schedule published in our FAQ; above that a run is scheduled individually. For us, the change described in this guide is a change of scale rather than a change of supplier — the board does not move buildings, and the documentation written at the prototype stage is the documentation the production run is built against.
Whether that is right for your product is a separate question, and it has an honest answer on the who we are right for page — including the cases where you should go elsewhere.
Why 100 boards cost so much per unit — stencils, setup, component minimums and test fixtures — and eight ways to bring the price down.
IATF 16949, ISO 9001 and ISO 13485 compared — and why a certification requirement is often what forces a supplier change.
MOQ, lead times by order quantity, component sourcing, certifications, board size limits and testing.
Gerber files, BOM and your assembly notes. We will come back within 24 hours with a price, a lead time, and a list of anything that would break the first production run — which is usually shorter than the list you were expecting.
