2D + 3D or you price the wrong part
Geometry is in the STEP file. Tolerances, finish and welds are only on the drawing. Read one without the other and you price a different part.
Sheet metal parts · Direct procurement
The price is decided on the drawing. Nobody has ever read them all.
A bend costs more than a hole. A tolerance nobody needs sends the part to a second machine. None of it is in the 3D file. Parsio reads the drawings too, rebuilds what each part should cost, and surfaces the references that no longer earn their price. Unlock 3 to 15% untapped savings.
Scope
Laser cutting Punching Bending & forming Welded assemblies Tube & profile Enclosures, frames & cabinets Powder coating & paint Galvanising & plating Machined metal parts Laser cutting Punching Bending & forming Welded assemblies Tube & profile Enclosures, frames & cabinets Powder coating & paint Galvanising & plating Machined metal parts
Why sheet metal is different
The cost is in the process, and the process is in the drawing.
What we learned sitting with buyers and design engineers in machinery, intralogistics and environmental equipment. None of it fits a generic sourcing tool, or a similarity score.
One bend is a second machine
Flat parts run thousands at a time off one laser. One bend moves the part to a press brake, with its own setup. The geometry barely changes. The cost does.
1 of 100 tolerances sets the route
Ninety-nine dimensions come straight off the laser. One tight callout sends the part to grinding, and decides the route and the quote.
1 project 1 new part number
A new project opens a new reference and a new drawing, even when the part was made two years ago. The portfolio grows. The volume per reference does not.
Savings plays
Six ways the model pays for itself.
Find the part you already buy
Cluster by what the part is: material, thickness, bends, weld class, finish. The same bracket redrawn for a new project stops being two references at two prices.
Price every part twice
A bottom-up should-cost from the process route, next to a predicted price regressed across your portfolio. The gap between them is the signal.
Challenge the tolerance, not the supplier
See which callouts force a process step, and what relaxing them is worth across every part that carries them. You arrive with the number, not the opinion.
Cost the part while it can still change
A new part is searched against the portfolio, then costed, before the drawing is frozen. The thickness set for the heaviest variant is not free on the lightest.
Match the part to the shop that can make it
Model suppliers by capability: which machines, which steps in-house, which subcontracted. Coating on site and coating trucked out are two prices and two lead times.
Price both sides of the variant trade
Engineering lowers unit cost by adding variants. Procurement loses more in volume, and the aftermarket loses again. Parsio numbers both sides.
How it works
From a folder of drawings to a costed route.
Read every file
Drawings, including the scans from before CAD. STEP files. Transactions and quotes. Geometry from the model, grade and tolerances from the drawing, fused into one record per reference.
Apply your rules
Your engineers write them in plain language: "a general tolerance is not a cost driver, a positional one on the laser datum is". Versioned, so when the engineer leaves the rule stays.
Model the shop floor
Cut, punch, bend, weld, treat, assemble, each with its own driver: metres of cut, bends, bead, coated area, nesting yield. Rates come from the supplier's own machine park, not a benchmark.
Compute both prices
Should-cost against predicted price, on every reference. Where they diverge, something is going on.
Customer outcomes
Outcomes, not features.
Save costs 14%
Taken out of a single-sourced product, priced from its own drawings
Cut complexity 4
Distinct cost families found inside what the ERP called one commodity
Move faster 1 week
From a folder of 400 drawings to a specified, costed portfolio
Worked examples
Steel container · Environmental equipment
One product, eight variants, one supplier, no idea.
Four BOM levels, about 150 parts per variant, one source, no way to tell if the price was right. Parsio read every drawing and STEP file, then rebuilt the supplier: machines, financing, subsidies, what is subcontracted. The model resolves down to a single bracket, and held up in the negotiation.
Welded assembly
~150parts per variant, four BOM levels
14%off the unit price
3suppliers rebuilt machine by machine
Built with sheet metal teams in
Machinery & machine toolsIntralogisticsEnvironmental equipmentAutomotive & commercial vehicles
Named references available under NDA.
FAQ
Questions, answered.
Your should-cost is higher than what we have been paying for years. Which one is wrong?
Often ours, and we will say so. A supplier holding a price for ten years is not losing money on it: there is a machine you have not modelled, a steel price you do not get, or a subsidy nobody mentioned. Every assumption in the model is visible and editable.
We have a cost engineering team and a should-cost tool already. What do you add?
The scenario, not the number. If you already compute should-cost, you know what a part costs on the route it takes today. What is missing is what happens when that route changes: relax this callout and the part comes off the laser instead of the grinder, which shops can quote it then, and what that is worth across every reference carrying the same callout. Parsio runs that across the portfolio rather than one part at a time.
The tolerance that matters is one line out of a hundred. How do you find it?
We extract all hundred, then rank them with your engineers. A general tolerance is noise. The positional one on the laser datum, or the callout that forces grinding, is the price. The ranking is your knowledge; reading every drawing is ours.
Our 3D files carry no tolerances and no finishes. Does that break it?
No, that is the normal case, and it is why we read both. Geometry from the STEP file, cost from the drawing. Parsio shows what it could not read before it shows a saving.
Some of our simple parts are expensive and some of our complex ones are cheap. Will the model be wrong?
On those, yes, and it will show you which. Usually a catalogue item sits inside the assembly, bought by your supplier and welded in, and nothing on the drawing says so. Parsio surfaces the outliers instead of averaging them away.
Do we have to replace SAP, Ariba or our sourcing tool?
No. Parsio is the layer before you decide an event is worth running. It exports the costed pack into whatever you already use. The ontology, the rules and the cost model are yours.
Bring one sheet metal family.
Brackets, enclosures, frames, weldments: a rough spend and whatever drawings you have, scans included. We'll tell you within a week whether it's worth a first phase, and we'll say so if it isn't.