Pricing a carbon part from its CAD geometry

In carbon composites, the hardest question a supplier gets asked is also the most common one: what will this part cost? The enquiry usually arrives as a CAD file and a deadline. Estimators answer it by experience and analogy. It's like the panel we did last year, but deeper, with an extra return flange. That works until the person who remembers last year's panel is on holiday, or the analogy turns out to be wrong and you discover the real margin after the job has shipped.

I built a tool that gives a first answer from the geometry itself. This is how it works, and what I'd tell anyone attempting the same thing.

The data was already there

The business had fully costed around ten vehicle programmes, roughly eight hundred parts each, and every one of those parts had geometry. The problem was never a lack of data. It was that nothing connected a new CAD file to that history. Each estimate started from a blank sheet and a memory.

Step one: read the geometry

STEP is the neutral exchange format almost every CAD system can write, which makes it the obvious input. From each file the tool extracts the features that actually drive effort in composite manufacture:

• Surface area, which drives material use and lay-up time.
• How sharply the surfaces curve. Tight radii make plies harder to drape, need more cuts and darts, and carry more risk of defects.
• Feature density: holes, flanges, returns and edges, each of which adds trimming and finishing.

Step two: one complexity score

Raw features are hard to compare. A large, gently curved panel and a small, tightly wrapped bracket can take a similar amount of work. So the features are combined into a single complexity rating, calibrated against what historic parts actually cost to make. Crucially, every historic part gets the same rating. That shared scale is what makes a new part comparable with the old ones.

Step three: price from history, not a formula

Parts of similar size and complexity take similar effort, so historic costs are the foundation. Current rates are then applied on top: labour, energy, which matters when curing is so energy-hungry, and material. Where the business has made the same part before, or something very like it, the historic base price becomes the anchor for the new quote. Bought-in operations are priced from real supplier costs, because in practice they are often the biggest variable.

Step four: show the margin before the quote leaves

The most important design decision had nothing to do with geometry. The tool calculates cost honestly and shows it next to the sale price and the margin. A loss you can see before quoting is a decision. A loss you find afterwards is a post-mortem. Estimators can override the material cost on any part, and every quote assumes production quantities rather than one-offs, because that is how the business actually builds.

Step five: keep people in the loop

Anyone can run an estimate, but a firm quote needs approval. The tool gives a fast, consistent first answer. The estimator's job shifts from arithmetic to judgement, which is where their experience is genuinely worth something.

What I'd tell anyone building one

• Start with the data you already have. Years of costed jobs are a training set most businesses never use.
• Make the answer checkable. Putting the historic base price next to a re-quote did more for trust than any accuracy figure.
• Keep cost and price separate. Mixing them hides exactly the losses you most need to see.
• Design for the reviewer, not the happy path. Overrides and approvals aren't a weakness in the model; they're how it earns its place.

The tool now covers composites, additive manufacturing, machining and fabrication, and it is used every day across commercial and engineering. The geometry was the interesting part to build. Making margin visible was the part that mattered.