There is a moment in every hardware program when the cost of being wrong stops being reversible. It is not the architecture review. It is not the EVT build. It is the day someone says cut the steel.
Everything after that day is cheaper to accept than to change. The mold exists. The fixture exists. The test sequence exists. From that point on, the company is not optimising a design — it is optimising a factory, and factories are very good at defending the decisions already baked into them.
Most founders arrive at that day later than they think, and less prepared than they need to be.
What tooling actually commits you to
A tool is not a purchase. It is a set of decisions compressed into metal:
- Geometry. Draft angles, wall thickness, shut-offs, parting lines. A design that looks elegant in CAD can be unmoldable, or moldable but only at a cycle time that destroys the unit economics.
- Tolerance. What the process can hold for pennies, and what it can hold for dollars, are different products.
- Volume assumption. A bridge tool at a few thousand units and a hardened tool at a million are not the same capital decision. Choosing one forecloses the other.
- The test strategy. A fixture designed at the same time as the enclosure catches defects. A fixture designed after the enclosure becomes a workaround.
Each of these is cheap to change in a CAD session and expensive to change in a tool shop. The whole job of NPI engineering is to move as many of those decisions as possible to the left — before the irreversible one.
The number nobody puts in the deck
A mold is not a large number in the context of a hardware company. That is precisely why it gets underweighted. The cost that matters is not the tool; it is the eight weeks after the tool, when the first shots come back out of tolerance and the schedule has to absorb a second cut.
Founders who have not been through it tend to model tooling as a line item. Founders who have been through it model it as the moment their unit economics became real. The bill of materials stops being a spreadsheet and starts being a contract.
What good looks like
The factories worth working with will push back before the tool is ordered. They will ask about draft, about gate location, about what happens to the part after it cools. They will tell you which of your cosmetic requirements are going to cost you yield. In a good review, the design changes and the schedule does not.
In a bad one, the quote is accepted, the tool is cut, and the first shots arrive with a note that says the design cannot be moulded as drawn — followed by a change order.
That difference is not a matter of factory quality alone. It is a matter of whether anyone with manufacturing judgment was in the room while the design was still soft.
The practical rule
If you are pre-tooling, spend the money on the review, not the tool. A DFM cycle is cheap. A second set of tools is not.
And if you are choosing between two factories, choose the one that argues with you earlier.