What does hardware manufacturing due diligence actually cover before a term sheet, beyond the pitch deck numbers?

Manufacturing due diligence examines five files you can physically check: the maturity of the design, the tooling and who owns it, the evidence that the process repeats, the chain underneath the bill of materials, and certification on parts coming out of the production tool. The numbers in the deck are the output of those five, not the proof behind them. Each file produces documents an independent engineer can read, and the absence of such a document is often the finding that matters most.

Most published guidance on this subject is written by advisors rather than by the party that builds the tools, which is why it tends to stop at the level of a questionnaire. The list below is organised around physical evidence instead: what exists, who owns it, and what it has been measured against.

What does the design tell you about manufacturability?

Whether it is frozen, and whether anyone has held it against a production process yet. Ask for the drawing set with revision numbers, plus the manufacturability report made against it, including which findings were accepted and which were deliberately declined. A declined finding is not a problem. A missing report means nobody asked the question.

Then count two things that rarely appear in a deck: the number of unique parts and the number of tools needed to make them. Every tool is its own build, its own correction loop and its own invoice. In practice a tool build runs six to twelve weeks, one correction round after the first trial is normal and two is unremarkable, and each round adds another two to four weeks. A product with twelve tools does not carry the same risk twelve times over; it carries twelve chances for the schedule to move, and they rarely move together.

Then look at the drawing itself, because the expensive assumptions live there. Tolerances that are effortless to draw in CAD cost money in series or cannot be held at all. Material specified only by family rather than by grade leaves room for a substitution that touches certification. And a colour that exists only in a render becomes a discussion the week the first parts come out of the tool. Research into design stage cost estimating puts roughly 70 to 80% of a product's life cycle cost in early design decisions, which is exactly why this file starts at the drawing and not at the spreadsheet.

What do you need to know about the tooling?

Three things: who owns it, whether the cavity count matches the plan volume, and whether it can be moved. All three live in a contract rather than on an invoice.

Ownership is the item that goes wrong most often and costs most to repair later. Paying for a tool does not automatically make the company its owner; the contract wording does. Where that wording is absent, the investment has not bought an asset but a production dependency, and a change of supplier arrives with a transfer charge nobody quoted. So ask for the clause naming the company as owner, with a documented right to move the tool when you change supplier, and ask where the tool physically sits.

Whether the tooling matches the plan is something you can calculate yourself. Cavity count times 3,600, divided by cycle time in seconds, times available running hours gives the theoretical maximum. A fictional example on invented assumptions: two cavities, a 25 second cycle and eighty running hours a week come out at just over 23,000 parts a week. Correct that for 10% downtime and 3% scrap and roughly 20,000 remain. Set that against the sales plan and you can see immediately whether there is headroom or whether a second tool belongs in the budget. Ask one thing alongside it: how many of those running hours are contractually the company's. A press shared with higher margin work is not capacity, it is an intention.

What evidence is there that the process repeats?

Measurement reports on parts from the production tool, followed by a run showing the process keeps hitting those dimensions. Those are two different things and they are often presented as one.

| What you request | What it proves | What a gap means | | --- | --- | --- | | Dimensional report on parts from every cavity | The tool can make a part within drawing | All that has been shown is that something comes out of the tool | | Capability run at production cycle time | The process repeats it | The first commercial run becomes the experiment, and that is not in the plan | | Scrap per cavity across three consecutive runs | The scrap figure in the cost model is real | Every percentage in the model is an assumption | | Inspection regime at an agreed AQL level | There is a rule for when a batch is rejected | Quality gets negotiated case by case, usually under time pressure | | Certificate on the production configuration | What is sold is what was tested | A design change during tooling correction may have invalidated the testing |

The industry standard for production parts, PPAP, commonly works with a capability index of 1.67 for a new process, 1.33 as a usual minimum for ongoing production, and a run of roughly three hundred consecutive parts at production cycle time. For batch sampling there is ISO 2859-1, where the AQL level fixes how many non conforming units a batch may contain before it is rejected. You do not have to prescribe those standards, but a company that knows none of them has already answered the question about how quality is controlled.

Check the provenance of every measured part. Parts from a 3D printer or a soft prototype tool say something about the design and nothing about the process that has to run the series. The same geometry out of a different tool is simply a different part.

Where does the chain risk sit underneath the BOM?

One layer below the first. A bill of materials names suppliers, but the risk usually sits deeper: with the party the supplier depends on, and with the single lane everything travels down. Four questions expose most of it.

• Which positions have a single source? Not just parts, but also a specific material grade, a colour, an electronic component or a single tool. Ask for a named alternative per position, with an estimated timeline and the cost of qualifying it. An alternative without a qualification path is not an alternative.
• What is the longest lead time in the chain? That drives working capital, not the average. Material and tooling are paid for months before the first sales invoice is collected.
• How many steps sit between raw material and final assembly, and how far apart are they? Distance in the chain costs freight, lead time and emissions at once. It is also the easiest lever to move while the design is not yet in steel.
• What compliance documentation exists per material? For the European market that means the declarations under REACH and RoHS, and the declaration of conformity for the product itself, with test reports on the configuration coming out of the production tool.

Name the trade off here as well, because a company that cannot explain it has probably not made it deliberately. Producing further away generally lowers piece price and raises lead time, inventory and the cost of every correction round. Producing closer reverses that. Which way a company has gone is not a mistake. Not knowing why is.

What does it mean when the file does not exist?

That you have a finding, not a rejection. In a young hardware company part of this file is almost always missing, and that is normal. The question is what it costs to obtain it, in weeks and in capital, and whether that time and money appear anywhere in the plan and in the manufacturing milestones each tranche is released against.

Three findings do belong before closing, because they are the most expensive to fix afterwards. Tooling ownership that is not on paper, because after closing you are negotiating about a tool that already exists. Tooling whose cavity count does not match the sales plan, because that means a second tool or a missed peak. And certification carried out on prototypes while the design changed afterwards, because then the paperwork exists and the cover does not.

The honest trade off in a thorough review: it costs weeks in a process where speed is itself a negotiating position, and it puts pressure on a company whose schedule depends on a toolmaker it does not control. So keep the scope small and physical. How those steps follow one another is set out in our process, and what a whole chain review looks at is described in the manufacturability review. On Higo Sense this ran across more than sixty parts from thirteen moulds with a one year certification deadline, and on HEARS it ran from prototype tooling through to production tooling for mass production. In both cases the risk did not sit in the spreadsheet but in the order in which the physical steps were taken.

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: capability indices of 1.67 for a new process and 1.33 for ongoing production, dimensional results per cavity and capability run size come from PPAP (AIAG), the widely used industry standard for production part approval. Batch sampling at an AQL level is defined in ISO 2859-1. The 70 to 80% of life cycle cost fixed in early design comes from Saravi, Newnes, Mileham and Goh (15th ISPE International Conference on Concurrent Engineering, 2008). REACH and RoHS are European regulations; the requirements per product follow from the current texts. Durations given are practical ranges rather than commitments, and the capacity example uses invented assumptions.