How do we estimate the cost of goods sold for a hardware product before committing capital, when the founding team has no manufacturing background?

Work the piece price back to the four manufacturing numbers that produce it: the number of cavities in the tool, the cycle time, the material used per part including the runner, and the scrap rate. If those four are not on paper, the unit cost in the deck is an assumption rather than an estimate, and it deserves a band measured in tens of percent that is skewed upward. A founding team with no manufacturing background cannot invent those numbers, but it can request them from the party that will build the tool and run the series. Requesting them is exactly what you can insist on before a term sheet, and it takes days rather than months.

Why is the unit cost in the deck rarely the unit cost?

Because what sits in the deck is almost always a piece price from a single quote, and a piece price is the narrowest number in the chain. It covers what that supplier offered: one part, out of his machine, at an assumed annual volume and an assumed batch size. Everything that happens outside that boundary is outside the number by default.

What routinely falls outside: packaging and labelling, inbound freight, import duty, outgoing quality control, secondary operations such as printing, assembly or ultrasonic welding, a scrap allowance, tooling amortisation, and the returns that come back after a year in the market. Each line looks small on its own. Together they move unit cost in most models by a double digit percentage, and always in the same direction.

There is a second and more fundamental reason. Research into cost estimating at the design stage keeps landing on the same order of magnitude: roughly 70 to 80% of a product's life cycle cost is fixed by decisions taken in early design. By the time the deck reaches you, most of the unit cost has already been put into the drawing. That makes the 3D model more interesting than the spreadsheet. The question is not whether the arithmetic adds up, but whether the geometry underneath it can be made in series at the number printed above it.

Volume deserves the same scrutiny. A piece price is valid at an assumed demand. If the sales plan halves, the piece price does not halve with it: tooling amortisation per part doubles, batches get smaller and changeover time is spread across fewer units. A cost model with only one volume scenario tells you nothing about the scenario an investor actually ends up in.

Which four numbers drive the piece price?

Cavities, cycle time, material usage and scrap. Together with the machine hour rate they account for almost all of the manufacturing cost of a moulded part. All four can be requested and all four can be checked.

| Number | Who supplies it | What it does to unit cost | | --- | --- | --- | | Cavity count in the tool | The toolmaker | Sets how many parts come out per shot, and therefore the machine time per part | | Cycle time in seconds | The toolmaker and the moulder | Multiplied by the hourly rate, this is the machine cost per shot | | Material per part, runner included | Follows from the 3D model and the runner system | Weight times price per kilo, and the runner counts until it goes back into the process | | Scrap rate as a percentage | The moulder | Inflates every other line: making 100 good parts means making more than 100 | | Machine hour rate | The moulder | Converts cycle time into money, and varies by machine size and by region |

Two of those numbers trade against each other, and that is where the judgement sits. More cavities cut machine time per part, but they make the tool larger, more expensive and more sensitive to fill imbalance. That trade only pays if the volume in the business plan actually arrives. An eight cavity tool for a product that sells 30,000 units a year is capital sitting idle in steel.

Cycle time is the number most often missing and the one that weighs most. Injection moulding research from RWTH Aachen puts 60 to 80% of the cycle in cooling, and cooling is built into the tool. A tooling quote without an expected cycle time is therefore not cheaper or more expensive than another, it is incomparable. Always ask for the tooling price and the piece price together, because separately they say nothing: a low tooling price can reappear in every piece price, and a low piece price can assume a tool nobody ordered.

How do you work that through to a landed unit cost?

By starting at the bottom, with machine time and material, and stacking the chain lines on top until the product is standing in your own warehouse. Below is that calculation for a fictional plastic housing part. The amounts are invented and exist to show the method; they say nothing about what a real product costs.

The assumptions: a polypropylene housing part of 60 grams with an 8 gram runner, two cavities, a 25 second cycle, a machine hour rate of 50 euros, a material price of 2.20 euros per kilo, 3% scrap, 150,000 units a year, and a 40,000 euro tool amortised over 600,000 parts.

| Line in the cost model | Per part | | --- | --- | | Machine time: 25 seconds across two cavities, at 50 euros an hour | € 0.17 | | Material: 68 grams including runner, at 2.20 euros per kilo | € 0.15 | | Scrap, 3% | € 0.01 | | Piece price as quoted | € 0.33 | | Packaging and labelling | € 0.05 | | Inbound freight to the warehouse | € 0.04 | | Outgoing quality control | € 0.02 | | Secondary operations: printing and assembly | € 0.06 | | Tooling amortisation: 40,000 euros across 600,000 parts | € 0.07 | | Landed unit cost | € 0.57 |

In this example the quoted piece price is 58% of the unit cost. The lines that are not in the quote add 73% on top of the number the deck is built on. That is the heart of it: the problem is not bad arithmetic, it is arithmetic drawn too narrowly.

Then move the assumptions around, because that is what you are underwriting. Seven extra seconds of cycle time, a normal outcome of one tooling correction round, adds roughly 5 cents a part here. If scrap runs at 8% through the first production runs, that is another cent or so. And if the sales plan delivers half the volume, tooling amortisation per part doubles. Those three together take unit cost in this example from 57 to about 70 cents, a quarter higher, without a single unreasonable assumption.

What accuracy band should the model carry?

One that narrows as physical evidence accumulates, written into the model rather than averaged away. Cost engineering has a standard frame for this: the estimate classification of AACE International, in which an estimate made without a developed design carries an accuracy of roughly minus 20 to minus 50% on the low side and plus 30 to plus 100% on the high side, while an estimate on a near complete design should land within a few percent. That classification was written for process plants rather than consumer hardware, so do not transplant the percentages literally. What transfers is the shape: the band belongs in the model, and it should narrow only when something has happened, not when something has been agreed.

For a hardware product there are three moments at which the band genuinely narrows. The first is a manufacturability review across the whole chain, putting geometry, material, assembly and packaging under the same light. The second is the point where two independent parties have quoted a tooling price and a piece price on exactly the same geometry and volume, with cycle time and a guaranteed shot count included; only then can you judge whether the cheapest tool is also the cheapest over the life of the product. The third is the first real shipment, when the invoice that arrives is reconciled line by line against the model. Only there does the model stop being a model.

One asymmetry is worth naming, because in hardware it almost always runs the same way. The lines missing from a cost model are additions, not subtractions. Packaging, freight, duty, inspection and secondary operations get added; there is rarely a forgotten line that lowers unit cost. So assume an incomplete model sits too low, and treat a model that lands exactly on the planned gross margin as a signal rather than a confirmation.

What do you ask for before committing capital?

Seven things, and all seven are a document or a number rather than an explanation in a meeting.

• The 3D model and a drawing carrying the dimensions that are functional, with a revision number.
• A manufacturability report against that revision, showing which findings were accepted and which were declined.
• A tooling quote and a piece price quote from two parties on the same geometry, with cavity count, steel grade, expected cycle time and guaranteed shot count.
• The scrap percentage built into the piece price, and who pays when actual scrap runs higher.
• The Incoterm the quote is based on, so it is clear how far freight is already included, plus the tariff code the product will be imported under.
• The full cost model with the chain lines in it, in at least two volume scenarios.
• The tooling contract, stating who owns the tool and whether it may be moved.

A team with no manufacturing background can request all of this; none of it requires experience. What such a team cannot do is judge whether the answers are plausible: whether 25 seconds is realistic at that wall thickness, whether two cavities fit that volume, whether 3% scrap is achievable in that material. That is where an independent review earns its place. The honest cost: putting two parties on the same geometry plus one manufacturability round usually takes a few weeks before the term sheet, and in a competitive round those weeks feel expensive. They are cheaper than the tranche that disappears into a tool the volume cannot support, and they belong to the question of which manufacturing milestones an investment tranche is released against. How those steps follow one another is set out in our process and in the manufacturability review across the whole chain. On Higo Sense that ran across more than sixty parts from thirteen moulds, with a one year certification deadline over the top of it.

Production Heroes takes products from design to series production, from feasibility and cost price to sourcing and supply chain, and supplies moulds and tooling to manufacturers across Europe, from the Netherlands and its own presence in China.

Sources: the 70 to 80% of life cycle cost fixed in early design comes from Saravi, Newnes, Mileham and Goh, "Estimating Cost at the Conceptual Design Stage to Optimize Design in terms of Performance and Cost" (15th ISPE International Conference on Concurrent Engineering, 2008); comparable figures are attributed in the literature to Ullman and to Rehman and Guenov. The share of cooling in the injection moulding cycle, 60 to 80%, comes from research at RWTH Aachen. The accuracy bands per estimate class come from the cost estimate classification of AACE International (recommended practice 18R-97), written for process plants. The worked example and the figures in the FAQ use invented assumptions and are not a quotation.