Moving From 100 Prototypes to 10,000 Units

Ten thousand is not a universal threshold. A moulded cover usually wants a tool well before that. A machined housing may still be sane there. The numbers in the title are a teaching range, not a BrahmWorks volume and not advice to tool on a calendar date. Every sum below is hypothetical teaching maths, not a price, a yield, a timeline or a test result. What a repeatable build has to contain is set out in How to Take a Hardware Prototype to Production.

A hundred units can be selected. Tight housings get paired with loose covers. A board that fails programming is set aside and not counted. The person who designed the cable route is in the room. Scrap stays a story, because the denominator is small and the exceptions are familiar.

Those habits do not scale by hiring. Selected fits become an undesigned sort. One programming failure becomes a lot quarantine. A cable route that lived in one person's hands becomes a missed station or a strain failure. Write down every selection before you multiply the order. Each is a design change, a fixtured step, or a loss you are about to buy at the larger quantity.

A shop that delivered a hundred brackets will not automatically hold price, capacity or finish at a much larger quantity, and it should not be asked to pretend to be a toolmaker. A broker who found odd reels is a poor plan for a year of production. Gratitude is not a supply strategy.

Unit cost changes shape. At low quantity, labour and machining dominate, and a tool looks like an insult. At higher quantity the tool spreads out, and the labour you stopped noticing dominates instead. Neither view is true at both ends. Compute both, with your quantity, before you call a process "too expensive" or "how we've always built it".

Cash moves earlier. A tool is paid before parts exist. A minimum order is paid before the units are sold. A hundred prototypes can be a project. The larger buy ties up material, and a wrong revision ties it up as scrap. A weekly CAD change, harmless at a hundred, orphans material at the larger buy.

You do not need a large quality department. You need lots that can be told apart, and a few checks that would have caught the last failure. Incoming checks on substitutions that still assemble. A station check before a cover hides the joint. A serial number tied to a bill revision and a firmware identity. A rule for what stops a lot.

Sampling does not replace a design that can be built wrong and still look finished. If a gasket can be rotated and look seated, a small sample will miss it. Fix the geometry. Inspect what geometry cannot make impossible, such as a bad reel or a missed torque.

Second sources belong on the parts whose absence stops you, and only after you have built both. A paper alternate is not capacity. At this teaching volume a second tool for a moulding may be unjustified, while a second qualified catalogue connector may be cheap insurance. Decide part by part. "Dual source everything" is a slogan that splits your volume and doubles your qualifications.

The teaching product is a connected instrument. It is not a BrahmWorks product. The first hundred used a machined aluminium shell, hand-soldered boards, and a cable made by a technician to a sample. They worked often enough to sell in small batches. The plan is to build about ten thousand over the following year, in the teaching scenario only.

What does not carry: sharp internal corners and milled hole bands, the technician's cable, the unwritten screw order, reflashing from a laptop. What can carry: the architecture, a stable board's function, and the user-facing requirements.

Teaching plan, not a BrahmWorks programme. A moulded or die-cast shell is quoted against a redesigned geometry, not against the machined file. The cable becomes a manufacturer part. Programming becomes a versioned fixture step with a logged identity. The first tooled lot is a pilot with dispositions, not an immediate jump to the full ten thousand. The hundred units already in the field keep their revision identity so a later fault is not blamed on a process they never saw.

Hypothetical teaching maths. Not a quotation and not a BrahmWorks result.

Assume a tool at INR 6,00,000. Spread over 100 units, that is 6,00,000 / 100 = INR 6,000 per unit, which swamps a small product and is a bad buy if you will never build more. Spread over 10,000 units, 6,00,000 / 10,000 = INR 60 per unit. The tool did not become cheaper. The quantity changed the question. If you might stop at a few hundred, do not buy the tool to look ready.

Assume hand work of 12 minutes per unit and an invented labour rate of INR 200 per hour. Labour per unit = 12/60 × 200 = INR 40. At 100 units, 40 rupees is not why you are late. At 10,000 units, minutes = 12 × 10,000 = 1,20,000 minutes = 2,000 hours. One person at about 1,800 hours a year does not clear that, before rework or holidays. The constraint is capacity, not the INR 40. A fixture that removes 4 minutes saves 4 × 10,000 / 60 = about 667 hours. At INR 200 an hour that is about INR 1,33,000 of the assumed rate, before you decide the fixture is worth its own cost. The hours are the point. A process that fitted in a prototype week does not fit in a production quarter just because the people are willing.

• Every hand selection in the hundred is a design change, a fixture step, or an accepted loss.
• Prototype suppliers are confirmed for capacity, or replaced on purpose.
• The geometry matches the process you will use at the higher quantity, not the prototype process.
• Tooling is spread across the quantity you will actually build, not the pitch.
• Cash for tools and minimum orders is planned before the purchase, not after.
• A revision freeze exists so material is not orphaned by weekly CAD changes.
• Serial, bill revision and firmware identity can be read back from a unit.
• Inspection is aimed at escapes geometry cannot prevent.
• Second sources are qualified or they are not listed.
• The first tooled quantity is a pilot with dispositions, not an assumption that scale is only arithmetic.

Only if you are confident you will build enough to spread it, and if the geometry is already the production geometry. A tool cut from prototype files, to celebrate a hundred working units, repeats the prototype in steel. If the next hundred may still change the split, the resin or the seal, stay on a short-run process and spend the money on learning. The INR 60 versus INR 6,000 illustration is the conversation, not a trigger.

It is enough to be hurt by a single point of failure on a common part. It is not automatically enough to pay for a second mould. Qualify a second source where the part is catalogue and the qualification is real. For a custom tool, a second source may be a second supplier who could cut a tool later, with the files in your control, rather than a tool you buy now and leave idle.

Some will. A shop that is honest about capacity, and a moulder who has already been in the DFM conversation, can. A supplier whose method was favours and night shifts will not survive a scheduled ten thousand. Ask for capacity and a repeatable process, and treat a reluctant answer as information. Moving is cheaper before they hold your only fixtures and your unreleased drawings.

Passes that depended on selection, on the designer, or on a process you are about to abandon. A radio result on the same board can carry if the board truly does not change. A drop result on a machined shell does not carry to a moulded shell of a different stiffness. A field anecdote from a hundred friendly users is not a defect rate. Keep the evidence that still refers to the configuration you will build. Retire the rest rather than cite it in a release note.

Bring the list of hand operations in the current hundred, the process you intend for the larger buy, and the quantity you will actually place. BrahmWorks can help separate what should be tooled, what should stay short-run, and what evidence has to be earned again.

Review the next stage with BrahmWorks