Data and machine learning
LiveCAD Geometry Cost Estimation
Upload a STEP file and receive a cost, derived from the part's geometry and years of historically costed work.
~8,000
Historically costed parts
10
Vehicle programmes of history
- Role
- Design, build and rollout
- Stack
- Next.jsPythonGeometry processingPostgres
The problem
New part enquiries were priced by experience and analogy. The business held cost data for thousands of parts it had already made, but nothing connected a new CAD file to that history, and margin was often only understood after the job.
The approach
- Geometry feature extraction from STEP files: surface area, curvature and feature density, combined into a complexity score.
- Pricing grounded in roughly 8,000 historically costed parts across ten vehicle programmes, plus labour, energy and material rates.
- Repeat and similar parts reference their historic base price, with supplier costs folded in.
- Separate workflows for composites, additive manufacturing, machining and fabrication.
- Cost, sale price and margin shown side by side, with approval required before any firm quote is issued.
The outcome
- Loss-making quotes are visible before they go out, not after the job.
- Used daily across commercial and engineering teams.
- Material overrides and approval gates keep estimates honest without slowing them down.
What they say
Mike built an estimating tool that reads the CAD and prices a part across carbon, print, machine shop and fabrication. The carbon model is fitted to 148 real parts we have made, so the number it gives is grounded in our own history rather than a rule of thumb. We see margin before a quote goes out instead of after.
Read the write-up
Pricing a carbon part from its CAD geometry
Want something like this for your business?
Get in touch