STEP File vs. Technical Drawing: Why Machine Shops Still Ask for 2D

You export a clean STEP file, upload it to a shop, and the first reply is not a price — it's a question: "Can you send the drawing?"If the model already describes every surface to a micron, what could a flat document possibly add? Quite a lot. The model describes the part's geometry. The drawing describes your requirements — and those are different things.

What a STEP file cannot say

A neutral CAD file is, in a machinist's words, "dumb geometry": mathematically perfect surfaces with no tolerances, no process notes, no priorities. Real machined parts are never nominal — every dimension lands somewhere in a range, and the entire game is agreeing on which ranges matter. Four things vanish in the export:

• Tolerances and fits. The model shows a ⌀20.000 hole. It cannot say whether ⌀20.05 is fine (clearance hole) or scrap (bearing seat). A press fit and a slip fit can be literally identical geometry.
• Threads. Most exports represent a tapped hole as a plain cylinder. M8 coarse or fine, thread class, tap depth — gone.
• Surface finish. Which faces need Ra 0.8 and which can stay as-machined is invisible in geometry.
• GD&T. Flatness of a sealing face, true position of a bolt pattern, runout on a shaft journal — the functional requirements that decide whether the assembly works.

What Xometry and Protolabs actually require

This is not folklore — the biggest manufacturing platforms state it in writing. Xometry requires a drawing for anything the model cannot capture: GD&T callouts, tap sizes, inserts, surface roughness finer than 125 µin Ra, and any tolerance tighter than ISO 2768 defaults. Protolabs Network says the same: threads, non-standard tolerances and non-standard finishes must be indicated in a technical drawing.

No drawing means the shop machines to general tolerance — and your bearing seat arrives as a clearance hole, legitimately, per the order. The drawing is not bureaucracy. It is the only mechanism that makes your requirements enforceable.

Why Model-Based Definition hasn't replaced drawings

In theory, MBD fixes all of this: embed the tolerances and annotations directly in the 3D model and retire the drawing. It has been promoted since the early 2000s. In practice, adoption stalled outside aerospace and automotive. In a NIST-cited industry survey, 85% of respondents used 3D models to carry product data — but only 41% used the model as the sole source.The majority still issue 2D drawings alongside, for unglamorous, rational reasons: suppliers' software can't read embedded annotations, inspection teams work from sheets, and the drawing remains the document of record in contracts.

What ambiguity costs

When requirements travel by phone call, email thread or guesswork, the bill arrives later as scrap and rework. APQC's cross-industry benchmark puts scrap and rework at 0.6% of revenue for top performers and 2.2% for bottom performers — and quality-cost analyses consistently find the true cost is 3–5× the visible direct cost once lost capacity, expediting, engineering time and missed deliveries are counted.

Against that, a one-page document that states tolerances unambiguously is the cheapest insurance in manufacturing.

The real bottleneck: making the drawing

So if drawings are this valuable, why do so many parts ship without one? Because producing a proper drawing has historically required a CAD seat, drafting-standards knowledge and hours of someone's time at $50–200 per hour. For a one-off spare or a simple bracket, the overhead often exceeds the cost of the part — so people skip the drawing and accept the risk.

That is the gap TechDraw AI closes: photograph the part, anchor one measured dimension, verify what matters, and get a drawing that meets the manufacturing-ready bar — without the drafting overhead that made skipping feel rational. The 2D drawing is not going away. Making one is about to stop being the bottleneck.