STEP, IGES, DXF, what each file actually gives an estimator

A buyer attaches a file and assumes the job is done. What arrives decides whether the next hour is estimating or drafting, and the 3 formats you will see most often are not interchangeable in the way the word CAD suggests.

STEP, IGES and DXF are the 3 you will see, and none is better in general. Each carries something specific, each drops something specific, and knowing which is which saves the request that gets quoted from the wrong file. Which one lands also decides how much of what AI can do for quoting applies.

What a STEP file gives you

A STEP carries solid geometry, which means the part has volume and you can ask it questions. Bounding box, mass given a material density, surface area, and the shape of every feature in space.

For a machined part that is the most useful of the 3. The stock envelope falls straight out of the bounding box, which sets the material line, and subtracting the part volume from the stock volume gives you the material being removed, which correlates with roughing time.

It also opens in almost anything. STEP is the neutral format that survives between CAD systems, which is why it is what most customers send when they send anything.

What it does not carry is anything about how the part is made or checked. No tolerances, no surface finish, no material specification, no notes. A STEP tells you the shape and is silent about everything that turns a shape into a price.

What IGES gives you and why it is usually worse

An IGES file carries surfaces rather than a solid, and that distinction is practical rather than academic. A surface model is a set of faces that may or may not enclose a volume, and frequently there are gaps.

If the surfaces do not stitch into a closed solid, you cannot ask it for a volume, which means no mass and no material removal figure. Somebody has to heal the model first, and that is 20 minutes of somebody who is not an estimator.

IGES is old and it survives because legacy systems export it. When it arrives it is often because the customer’s CAD is old, which usually correlates with a legacy part, no current model, and a drawing that is itself a scan.

IGES arrives with a story attached.

Treat an IGES as a signal as much as a file. It is worth asking whether a STEP exists, because if it does the request takes 2 minutes and saves the healing, and if it does not you have learned something about the part’s history.

What a DXF gives you, and why it is the right file for sheet work

A DXF is 2D and that is exactly what a sheet metal part needs. It carries the flat profile, which is the blank, which is the thing you nest and the thing you cut.

For a sheet metal quote that is more useful than a solid. Cut length comes directly from the profile, internal features are countable, and the nesting yield that dominates the material line can be computed properly rather than estimated.

What a DXF does not tell you is which lines are bends. Some carry a bend layer and many do not, so the bend count, the direction and the deduction all have to come from the drawing or from the 3D file.

It also has no thickness, no material and no grain direction. A flat profile with no gauge is not a quotable part, which is why a DXF alone is a common and frustrating attachment.

The blind spot all 3 share

None of these formats carries tolerance, finish or material specification. That is not a limitation of any one of them. It is what a drawing is for, and it is why a model without a print is not a quotable package.

The cost consequence is large. A bore at a general tolerance and the same bore at half a thousandth are different operations with different machines and different inspection, and the geometry is identical in both cases.

Surface finish is the same. A face called at 125 and a face called at 16 look the same in a STEP and are a grind apart in the routing. Heat treat and plating appear only as notes.

Where GD&T is involved the gap widens further, because position and profile controls change fixturing and inspection in ways that have no geometric expression at all. A model gives you the nominal and says nothing about what has to be held.

Geometry is the cheap half.

There is a format that closes this gap and it is worth knowing about even though it rarely arrives. Model based definition carries tolerance and annotation inside the 3D file, so the model becomes the controlled document. Some aerospace and automotive customers work this way. Most do not, and the practical consequence is that a drawing is still the thing that tells you what the part costs.

A quick way to judge what landed

Open it and ask 3 questions before anything else. Does it have a volume, which separates a solid from a surface set. Does it match the drawing revision. And does the drawing exist at all.

Those 3 take a minute and they route the request. A closed solid matching a current drawing is a package you can quote today. A surface set with no drawing is a conversation with the customer before an estimator touches it.

The answer also tells you how much of the work is drafting rather than estimating, which is worth knowing before somebody promises a date.

Open it before you promise a date.

What to do with what actually arrives

Match the ask to what is missing rather than asking for everything. If a STEP arrived with no drawing, ask for the drawing. If a DXF arrived, ask for the thickness, the material and whether the bend lines are on a layer.

If nothing but a scan arrived, that is its own exercise and the model request should go out in the first hour rather than the last.

Where a model and a drawing disagree, the drawing governs for quoting. Models get updated without the drawing being reissued more often than the reverse, but the drawing is the controlled document and it is what a first article will be measured against.

Record which file you quoted from and its revision, the same way a takeoff from a scan should. When the customer sends a new model 8 months later, knowing what you priced against is the difference between a quick comparison and starting again.

The last thing worth checking is that the file matches the drawing at all. A model at a different revision from the print is common on legacy programs, and the tell is usually a dimension that does not agree. Finding that at quote time is a question. Finding it at first article is a drop of material and a schedule problem.