CNC prototype essentials: what decides a first article
A working guide to the few decisions that actually change a CNC prototype: stock removal strategy, datum choice, tolerance zones and finish. Written for design and manufacturing engineers who need a functional part on the bench, not a cosmetic model.

In this article
- 1
- 2
- 3
- 4
- 5
- 6
Why a CNC prototype behaves differently from a drawing
A CNC prototype is subtractive from the first cut. The tool has to reach the surface, the chips have to leave, and the part has to stay put. Those three conditions shape the geometry more than the drawing does. A pocket 40 mm deep and 8 mm wide is legal on paper. In aluminium it needs a long-reach tool that deflects, so the wall tapers and the floor chatters.
Climb milling on a rigid setup leaves a predictable surface. Conventional milling tends to smear aluminium and work-hardens stainless. That is why the same drawing can come back with Ra 0.8–1.6 μm from one shop and Ra 3.2 μm from another. The machine did not change. The feed direction and the tool engagement did.
Heat is the quiet variable. Cutting 7075 aluminium at 12,000 rpm without enough coolant raises the part temperature by 15–25 °C. By the time the part cools to the gauge, a 200 mm length has moved. Titanium TC4 moves the other way: it conducts poorly, so the edge takes the heat and the insert wears faster than the part grows.
None of this means tight parts are impossible. It means the first article carries the accumulated error of every setup, tool change and thermal cycle. A prototype drawing with ±0.005 mm on a feature that crosses three setups is asking for a rework loop. A drawing that keeps critical features in one setup usually passes on the first pass.
- 1Reach beats toleranceIf the tool cannot reach it, the tolerance does not matter.
- 2One setup, one datumFeatures machined without re-clamping hold position best.
- 3Heat moves metalLet the part stabilize before the final cut and the final gauge.
CNC prototype essentials: datum choice and setup count
Every re-clamp adds a new zero. On a 3-axis machine a five-sided part needs four or five setups, and each one stacks its own error. On a simultaneous 5-axis center the same part can often be finished in two. The table rotates, the tool stays normal to the surface, and the datum never moves. That single difference is the main reason 5-axis work is not only for complex shapes.
Choose datums that exist on the raw stock, not on a surface that will be cut away. A face milled flat on the first operation makes a good secondary datum. A cast or saw-cut face does not. If the drawing calls out a datum on a surface that only exists after operation three, the inspector and the machinist will measure from different places.
Design for one open side where possible. A part with all critical bores reachable from the top can be finished in one setup, then flipped only for deburring. A part with bores on four faces will need fixtures, and fixtures add cost and lead time. On a prototype run of one to twenty pieces, that cost is real.
Plan the order of operations around the features that must stay concentric. Bore it, then turn the mating diameter in the same grip if the machine is a mill-turn center. Our mill-turn centers hold that relationship without a second zero. When a prototype must be turned and milled on separate machines, add a dowel-pin fixture instead of trusting the chuck.
- 1Stock datums firstKeep at least one datum on the as-received blank.
- 2Fewer faces, fewer zerosCritical features on one face cut setup error sharply.
- 3Mill-turn for concentricityTurn and mill relationships survive in one grip.
Tolerances and surface finish that a prototype can actually hold
±0.005 mm is achievable on a milled feature in aluminium with a rigid setup and a temperature-controlled room. It is not achievable on a 400 mm thin wall, and it is not achievable on a deep pocket with a long tool. The tolerance belongs to the feature, not to the shop. Put the tight callout only where the assembly needs it.
Position tolerance is usually the better control on hole patterns. A hole-to-hole tolerance of ±0.005 mm on a 300 mm bolt circle is harder than a true position of Ø0.05 mm at MMC, because the second one allows the bonus tolerance from the hole size. Engineers who switch to position callouts on prototypes usually get fewer inspection arguments.
Surface finish follows feed and tool radius. A 0.4 mm corner radius on a ball-nose tool leaves a scallop height that sets the floor finish. If the drawing asks for Ra 0.2–0.8 μm on a curved surface, the shop has to step over at a very small increment, which multiplies cycle time. Ask for Ra 0.8–1.6 μm on the as-machined surface and polish only the sealing face.
Anodizing and plating change dimensions. Type II clear anodize adds roughly 5–12 μm per surface. Hardcoat adds more and grows corners. If a bore must stay within ±0.005 mm after coating, mask it or machine it undersize on purpose. Tell the shop before the finish, not after.
- 1Tight only where it mattersApply ±0.005 mm to the mating feature, not the whole part.
- 2Position over plus/minusPosition with MMC gives the shop usable room.
- 3Finish changes sizeMask coated bores or pre-size them under the limit.
Material choice for a functional prototype
Match the prototype material to what the test is measuring. If the test is fit and function at room temperature, 6061-T6 aluminium is fast, stable and cheap to machine. If the test involves wear, 7075 or 17-4PH stainless holds a better edge and better strength. If the test is corrosion or a medical clean, 316L and 316 are the right starting point.
Plastic prototypes have their own rules. POM machines cleanly and holds tolerance. PEEK needs sharp tooling and a slower feed or it burns. Carbon-fibre composite is abrasive, so expect tool wear and a rougher edge. ABS and PC are fine for housings but move with moisture, so gauge them dry if the tolerance is tight.
Thin walls are where material choice bites. A 0.5 mm wall in aluminium will deflect under light clamping. The same wall in stainless is stiffer but harder to cut without work hardening. In plastics, a 0.8 mm wall in POM is workable; in ABS it will bow. If the prototype has a thin wall, add stiffening ribs or accept a looser tolerance.
Do not switch material between the prototype and the production part without re-checking the fits. Aluminium to stainless changes thermal expansion and thread strength. A plastic prototype that snaps together may not represent a die-cast housing. The prototype is only useful if it tests the same failure mode.
- 1Test first, then materialPick the alloy that reproduces the failure you are studying.
- 2Thin walls flexBelow 0.8 mm in aluminium, expect clamp marks and bow.
- 3Keep the failure modeA change of material can hide the problem you are testing.
What to check before releasing the CAD
Run a quick self-check before the file leaves your desk. Confirm that every tool can reach its surface. Check the minimum internal corner radius against the largest tool that fits the pocket. Look for deep, narrow slots. A slot 6 mm wide and 60 mm deep has a 10:1 depth-to-diameter ratio, and that is where chatter starts.
Check the thread callouts. A 1/4-20 thread needs a clearance hole of at least 6.8 mm. An M3 tapped hole in aluminium wants a 2.5 mm drill and at least 6 mm of full thread engagement. Threads that run too close to a wall bulge the wall. Threads under a boss barely fit a tap wrench.
Add chamfers or radii to inside corners. A sharp internal corner is a stress riser and a tool breaker. A 0.5 mm corner radius is enough to help, and it does not change the fit. On parts that will be anodized, break all edges by 0.2–0.3 mm so the coating does not bridge.
Send the native CAD plus a STEP file. STEP carries the geometry, but the native file helps when a face needs to be rebuilt for a 5-axis toolpath. If you can only send one, send STEP AP242 and mark the critical features on a 2D drawing. A marked-up drawing prevents the most common first-article miss: a tolerance applied to the wrong side of a fit.
- 1Depth-to-diameter under 6:1Deeper slots need a smaller tool and slower feeds.
- 2Break the sharp cornersA 0.5 mm inside radius reduces chatter and stress.
- 3Send native and STEPBoth files together shorten the review cycle.
When a feature suits 3-axis, 5-axis, or needs a design change
Use this as a screening table before you release the prototype drawing.
| Feature | Best process | Practical limit | When to redesign |
|---|---|---|---|
| Open pocket, one face | 3-axis milling | Depth-to-diameter under 6:1 | Slot deeper than 8× width |
| Five-sided housing | 5-axis simultaneous | Two setups instead of five | Under-cut deeper than tool reach |
| Concentric bore and face | Mill-turn center | Same-grip runout under ±0.01 mm | Separate machines without a fixture |
| Thin wall under 1 mm | 3-axis with soft jaws | Wall bow 0.05–0.15 mm | Add ribs or thicken to 1.5 mm |
| Hole pattern on a bolt circle | 3-axis, position callout | Ø0.05 mm true position at MMC | Keep plus/minus only if needed |
| Sealing face, Ra 0.2–0.8 μm | 3-axis then lap | Polish only the seal band | Full-face spec raises cycle time |
| Coated bore, tight fit | Mask before anodize | Growth 5–12 μm per surface | Pre-size under the lower limit |
| Prototype in PEEK or Ti | 5-axis, slow feed | Tool wear changes the last part | Split into two stable parts |
The takeaway
If your prototype has critical features on one face and a single datum, a 3-axis machine with a good fixture will hold ±0.005 mm and ship faster. If the feature crosses four faces or has an under-cut, choose 5-axis and accept the higher hourly rate. Redesigning to avoid a five-setup part is usually cheaper than machining it.
Questions engineers ask before the first cut
What tolerance can a CNC prototype realistically hold?
On aluminium with a rigid setup, ±0.005 mm is achievable on a feature that is machined in one setup. Position tolerance of Ø0.05 mm at MMC is a practical callout for hole patterns.
Deep pockets, thin walls and long tools lose accuracy. On a 400 mm thin wall, expect 0.05–0.15 mm of movement even with light clamping. Put the tight callout only on the mating feature.
How long does a prototype take from file to part?
Quotation and a free DFM analysis come back within 12 hours. Production can start within 24 hours of approval, and parts ship in 3–5 days for most geometry.
Five-axis parts with complex toolpaths and coated surfaces take the longer end of that window. Parts that need a custom fixture add a setup day.
Is there a minimum order quantity for a prototype?
No. We run from one piece to runs of 10,000 or more. A single first article goes through the same inspection route as a production batch.
Every part is inspected before shipment, with raw-material check, in-process monitoring and final inspection. Reports are available on request.
Which file format should I send?
Send STEP AP242 plus the native CAD file, and mark critical features on a 2D drawing. STEP carries the geometry; the native file lets us rebuild a face for a 5-axis toolpath without guessing.
If the part has a tolerance stack across several features, include the stack calculation. It tells the machinist which dimension to hit first.
Will anodizing change my dimensions?
Yes. Type II clear anodize adds roughly 5–12 μm per surface. Hardcoat adds more and builds on corners. A bore specified at ±0.005 mm will not stay there after coating unless it is masked.
Decide before machining. Masking a bore or pre-sizing it undersize are both workable, but neither can be added after the part is finished.
How do you protect the design?
Uploads are handled as secure and confidential. An NDA is available on request, and we can work under yours.
We hold ISO 27001:2022 for information security, alongside ISO 9001:2015, IATF 16949:2016 and ISO 13485:2016.
Send the drawing, get a DFM review with the quote
Upload the CAD and the marked-up drawing. Within 12 hours you get a price, a setup plan and a short list of features that will not hold the tolerance as drawn.
12-hour quoteFree DFM analysisNo MOQ100% inspection