A quote that comes out too high, you lose. A shame, but clean: the order passes you by and that is the end of it. A quote that comes out too low, you win. The customer signs remarkably fast, the shop floor gets to work, and only when the actual hours come in does it turn out this was an order you would rather have lost. Estimating on gut feeling quietly organizes its own adverse selection: you lose the sharp orders and win the loss-making ones.
The good news: the numbers that make an estimate solid already exist. Every Inventor model contains weights, materials, dimensions and a bill of materials, exactly the data that material and processing costs rest on. Cost estimation on CAD data therefore does not require a new software package, but a route to get that data into your quote without retyping, plus the data quality to dare to build on it. This article walks that route step by step, including the role a toolbox like Thundercad plays in it.
Why gut feeling wins the wrong orders
Every estimate has noise. An experienced estimator lands above the mark sometimes and below it other times, and over many quotes that averages out nicely, you would think. But the market does not average along. The quotes where you were too high disappear without a sound; the quotes where you were too low come back as orders. Your order book slowly fills itself with your own estimating errors.
Concretely, with assumptions: suppose a frame is estimated at 24 hours of welding and finishing, while in reality it takes 31. Once is bad luck. If that underestimate is structural, you win exactly the frames your competitors were sensibly too expensive for. The remedy is not thinking harder, but estimating less: trace as many cost items as possible back to measurable data from the model, and write down the rest explicitly as assumptions.
What your model already knows
A sizable part of your cost price hangs on data the model already contains or can produce in one step:
| Model data | Cost item | Example from practice |
|---|---|---|
| Weight and material per part | Material purchasing | Kilos of structural steel for the frame, kilos of stainless for the cover |
| Flat patterns and sheet dimensions | Sheet usage and cutting | Nesting on standard sheet, estimating remnant material |
| Surface area | Blasting, coating, galvanizing | The square meters the coater bases his price on |
| Quantities in the BOM | Processing hours | Sawing and drilling twelve identical supports goes faster per piece than one |
| Purchased parts in the BOM | Supplier quotes | Motors, bearings and chains ready to request from the supplier |
The pattern: the more items you trace back to measurable model data, the smaller the part that floats on experience. The processing hours remain an estimate, but an estimate per operation based on real quantities and dimensions, instead of one big total at the bottom of the page.
From BOM to costing sheet without retyping
The classic route between model and estimate is retyping: the bill of materials on one screen, Excel on the other. That easily costs an hour per quote and introduces exactly the errors you wanted to avoid: a swapped line, a missed part, a weight from before the latest change.
The better route is exporting into a fixed template. With Export BOM, Thundercad puts the bill of materials of an assembly into your own Excel layout in one go, with the columns your estimate needs: article number, description, material, weight, quantity, make or buy. Your costing sheet references those lines and works out the material items by itself; the estimator keeps his head free for the items that genuinely require judgment. Do refresh the BOM before exporting, so this morning's change is included.
The same export serves more purposes than the quote alone, by the way. Work preparation uses the same overview for ordering material, and on a repeat order you quickly put the new BOM in Excel next to the previous one. One fixed route from model to Excel pays off in more places than just the quoting table.
Curious how fast your bill of materials lands in your own costing template? Export BOM works with your existing Excel layout, so your costing sheet does not need an overhaul.
Try 30 days freeData quality first, building second
An estimate on polluted model data is more dangerous than gut feeling, because it looks trustworthy. A part without a material counts as zero kilos. A wrong density makes the cover half its real weight. An outdated BOM misses the reinforcements from the latest change. Before you build your quote on the model, a short checklist has to be green:
- Every part has a real material, not a default that was never replaced.
- No zero weights in the BOM: every zero weight is either an error or a deliberate exception with a note.
- The make-or-buy field is filled in everywhere, so your costing sheet sorts the lines correctly.
- The BOM has been updated after the latest model change.
Also agree on who delivers what. Engineering owns the model data and delivers a BOM with this checklist ticked off; the estimator gets to trust it without re-measuring everything. If the estimator does find a gap, that report goes back to engineering and the source gets fixed, not just the costing sheet of that one quote.
How to get weights and materials structurally right, and keep them right, is something we worked out earlier in Weights and materials in your BOM that are actually right; consider that article the foundation under this one. And do not underestimate the flip side: the same data errors that skew your quote surface later as rework on the shop floor. How to trace that damage back to its source is covered in Failure costs from data errors: follow them from the shop floor back to the source.
Actual hours: closing the learning loop
Estimating on model data only becomes truly strong when you measure the outcomes. Compare the estimated hours with the actual hours for every finished order, split by operation type: sawing, bending, welding, assembly. Not to grade anyone, but to adjust standard times.
An example, explicitly with assumed numbers: if welding the platforms turns out to take roughly a fifth more hours than estimated three orders in a row, that standard time is too optimistic and you adjust it. Order by order, each product family builds up a set of standard times that keeps improving, and the discussion at the quoting table shifts from "what do you think?" to "what do the last orders say?".
Keep the ritual small, or it dies. Fifteen minutes per finished order is enough: put the actual hours next to the estimated ones, flag deviations above an agreed threshold, and only adjust a standard time after a few orders show the same picture. Make it any bigger and nobody keeps it up past the third order.
Frequently asked questions
How detailed does the model need to be at quoting time?
Less than you think. A rough model with the correct main dimensions, real materials and the major purchased parts already puts the material items on measurable ground. Detailing follows after the order; the estimate works with what is there and explicitly marks the rest as assumptions.
Does estimating on model data work for one-offs and specials?
Especially there. In series production, repetition corrects your estimating errors by itself; with one-offs there is no second chance and no history to lean on. So estimate specials module by module, and lean where possible on comparable modules from earlier orders.
Where do I start without making it a big project?
Pick one product family and build a costing sheet for it that runs on exported BOMs. Run a few quotes in parallel, the old way and the new way, and compare them when the actual hours come in. If you want to test the export right away, try Thundercad 30 days for free.