Hardware Product Development: From Idea to Mass Production

Nothing below is a BrahmWorks test result, a client cost or a supplier quotation. The arithmetic is teaching maths on invented inputs.

A buildable idea is not a product. Hardware ties cash up in parts, tools and people, and a bench unit does not create a buyer. Before industrial design, electronics or firmware start, write five lines a sceptical partner can reject.

Name who pays, and the substitute they pay for now. If you cannot, you are exploring, not developing. Name one requirement you would not ship without. If every miss is “we will see”, the idea is not specified. Name the quantity the business needs, and the process that quantity can support. A consumer price that only a machined enclosure can hit is a contradiction. Name who services it. A sealed unit you cannot open is a consumable: price a replacement, not a repair you will never do. Unproven physics or materials are research, with their own stop, not a task hidden in the product schedule.

If those lines conflict and nobody will change one of them, stop. A short experiment that could disprove the idea is cheaper than a polished prototype of the wrong product. Write the rejection down, or a later demo will pull the dropped scope back in.

Once an idea survives, the danger moves from “should we build anything” to “which statement, if wrong, wastes the next sum of money”. A requirement is a risk when it is cheap to type and expensive to prove, or when it shares a part with another requirement.

Sort each one into four classes. Safety: someone can be hurt. Legal sale: you cannot place it on the market you named. Function: the buyer rejects it. Preference: it is less liked, but still legal and usable. Safety and legal sale are not items to trade with a probability sum. They need evidence and an owner who can stop the release. Function can be retired with a test. Preference can wait.

Look for coupling. A seal, a radio, a cell and a drop rating often sit on one enclosure, so four lines are one geometry. Decide which requirement wins when they conflict, or the last CAD edit will.

For each function risk, name the cheapest honest check that could kill it, and the stage at which a miss becomes a scrapped tool or a wasted submission rather than a bench change. Record the check, the configuration and the result that would make you stop. A requirement with no method and no owner is a wish.

Certification is not a stamp you buy when the design feels finished. It is a choice of claims and markets, and therefore of the evidence the design must produce. That choice belongs with the idea, because it changes the product.

List the markets in the first production run, and the markets you will not enter. List the claims you will make — mains, radio, outdoor, medical, child use, a regulated measurement — and the claims you will not make. Dropping a claim can remove a body of evidence and can also remove customers. Write both. A person qualified for that market should name the regime that follows. Do not take it from a blog, a competitor’s label, or a factory’s assurance that “this usually passes”.

The scope changes the geometry. Creepage, antenna placement, enclosure material, markings, the manual and the ports move when the claim moves. Adding a market after the tool is cut can mean a new tool and a new test. Ten thousand units in two regulated markets is a different programme from 100 units under a narrow claim.

Before you freeze geometry, ask the lab or compliance partner what must be representative of production, which change voids the result, how many samples they need, and who pays for a resubmission. Record the answers against a revision. “We will certify later” discovers the path when change is most expensive.

A pre-compliance check is not a certificate, and a pass on a different enclosure does not travel. Formal evidence has its own invoice. Keep it off the engineering fee so a repeat is visible.

Mass production is not a phrase. Each quantity asks a different question. Use the next column too early and you freeze an immature design. Stay in the earlier column too long and you cannot defend the cost or the quality.

Fund the column the evidence supports. A plan written for 10,000 units, while you only have 10, should pay for the next column, not the tool for the last one. An unqualified cheaper part at the high quantity costs more than it saves when it fails in the field or in a surveillance check.

This calculation is invented so the method is visible. It is not a quotation, not an Indian market rate, and not a yield BrahmWorks has measured.

Assume you need saleable units, not starts. A failed unit is scrap, with no rework. Variable cost is INR 650 per unit started, parts and assembly only. A fixture costs INR 1,80,000. You assume — you do not know — that it lifts first-pass yield from 85 percent to 96 percent. Ignore tax, freight, debug labour, warranty, and the chance the fixture does nothing.

Scrap units = saleable × (1 − yield) / yield. Scrap cash = scrap units × 650.

For 1,000 saleable units at 85 percent, scrap units = 1,000 × 0.15 / 0.85 = 176.47, and scrap cash = INR 1,14,706. At 96 percent, scrap units = 1,000 × 0.04 / 0.96 = 41.67, and scrap cash = INR 27,083. The fixture saves INR 87,623 of scrap and costs INR 1,80,000, so on this batch it is INR 92,377 short if scrap is the only benefit.

For 10,000 saleable units the same rates scale by ten. Scrap cash falls from INR 11,47,059 to INR 2,70,833, a saving of INR 8,76,226. Against INR 1,80,000 the surplus is INR 6,96,226, still only under these invented rates.

Repeat the arithmetic with a quoted fixture price, a quoted variable cost and a yield change you will stand behind. If you cannot name that yield change, you do not have a reason to buy the fixture. At 10 saleable units the same 85 percent scrap is about INR 1,150. Do not let a 10,000-unit story choose the tool for a 10-unit fact.

A supplier is ready when the process that will make the order can hold the characteristics you marked critical. A hand-dressed sample does not show that. Ask what they will not do: the resin, the tolerance, the lot size, the test, the packaging. A marked-up drawing is a stronger signal than a polite yes to the whole file.

Match the ask to the column. At 10 units, a shop that will make two pieces is enough. At 100, the same process has to be repeatable enough that a defect cause means something. At 1,000, you need incoming rules, a named approver for deviations and orderable part numbers. At 10,000, you need capacity, a second source for a part that can stop the line, and a rule for tool maintenance. A missing second source is noise at 100 and a business risk at 10,000.

Write down tool ownership and where the tool sits if you move, the minimum order, the quoted lead time, who approves a deviation, and who pays for a rejected lot. A low piece price with a large minimum spends cash before you have field evidence.

You still owe materials, critical dimensions, a definition of good, and a person who answers a deviation in days. Once the design functions and the question is repeatable manufacture, that package is the work in How to Take a Hardware Prototype to Production. A factory named before the package exists is not a ready supplier.

Each line needs an owner and a place the evidence lives. A tick with no record will not survive the next quantity.

• The paying user, today’s substitute, and one requirement you would not ship without, are written.
• The business quantity, the process and the service story match. A unit you cannot open is priced as a consumable.
• Requirements are classed as safety, legal sale, function or preference.
• Coupled requirements that share a part name a winner when they conflict.
• Markets in, markets out, claims in and claims out are on one page.
• A qualified person has named the regime, and what change voids the evidence.
• You are funding the column you are in: 10, 100, 1,000 or 10,000.
• Any tool has a quantity where its saving appears, and the yield change is labelled as an assumption.
• The supplier has shown a process, not only a sample, for the critical characteristics.
• Tool ownership, minimum order, lead time and deviation approval are in writing.
• Repeatable build and test is either not yet due, or it is the scoped work in the article linked above.

You can name a payer, a substitute, a requirement you would not ship without, a quantity whose process matches the price, and a service story you can perform. Unproven science is a research stop, not a hidden task. Wanting a prototype is not a test.

When the claims and the first markets are chosen, before geometry and tools are frozen. Book the lab when the design is representative. Do not leave the claim until after the tool is cut. A new claim later is usually a new design.

They are a repeatable batch if the process, fixtures and incoming checks exist and another order is affordable. They are not a rate. Ten thousand adds capacity, change control and the cost of a stop. Ten units are still samples.

You know the quantity column, the critical characteristics, and who owns tools and deviations. The manufacturer has marked what they will not do. If the open question is repeatable build and test, hand over a package, not a verbal brief.

Bring the idea filter, the ranked requirements, the claims and the quantity you intend to fund. BrahmWorks can say which column you are in, and whether the next spend is a learning build, a supplier trial or repeatable production.

Review the next stage with BrahmWorks