Guide
What drives CNC machining cost on a custom part?
CNC machining cost is set mostly by decisions made before the first cut: the material and stock form, how many times the part is re-fixtured, how long it spends in the cutting cycle, how tight the tolerance and finish call-outs are, and how many pieces are ordered at once. None of those appears as its own line on a quote, but together they are most of it.
By Saola Manufacturing engineering team · Published · Updated
This guide walks through each of those drivers and the drawing or RFQ change that typically brings it down. It does not publish a price: material markets, machine rates and labor cost move by shop, by material and by month, so the only reliable number for a specific part comes from a quote against the actual drawing. What follows explains why that number is what it is, and which choices made earlier in the design move it before the drawing ever reaches a shop floor.
On this page
What are the main drivers of CNC machining cost on a custom part?
Cost is set by how much material has to be removed, how many times the part is set up in a machine, how long the cutting cycle runs, and how much inspection the tolerance and finish call-outs demand. Order quantity then decides how those fixed costs are spread across the parts that share them.
The table below lines up each driver against the drawing or RFQ decision that moves it. The sections that follow take each row in turn, in more depth than a table allows.
| Cost driver | What moves it | Drawing or RFQ lever |
|---|---|---|
| Material & stock form | Machinability of the alloy, and whether the size matches a stocked bar, tube or plate | Specify a material and section size the shop already runs |
| Setups & fixturing | Every re-orientation of the part needs its own alignment and, often, its own fixture | Design features reachable from as few sides as possible |
| Order quantity | Setup, programming and gauging cost is fixed per run and spreads over however many parts share it | Batch similar parts together; state the realistic order range in the RFQ |
| Cycle time | Feature count, deep pockets, small internal radii and thin walls all slow the cut | Open radii to standard cutter sizes; keep walls thick enough to resist clamping load |
| Tolerance band | A band narrower than the process needs slows finishing passes and increases how often parts are checked | Hold tight tolerance only on features that mate, seal or locate |
| Surface finish & masking | A fine roughness call-out can add a pass; a coating adds a masking step on surfaces that must stay bare | Call out finish only where a seal or bearing lands; mark what must stay bare |
| Threads | A standard thread uses gauges a shop already owns; a custom one needs its own gauge set | Specify a standard thread form and class where the design allows it |
| Inspection plan | Sampling density and the instrument a tolerance forces, from a hand gauge to a coordinate-measuring machine | List which features need a full dimensional report and which follow the general note |
| Casting vs. machining from solid | Casting removes stock-removal time but adds tooling cost and time to first parts | Model both routes at the quantity actually expected |
How much does material choice and stock form change the cost?
Machinability sets how fast the material can be cut and how much tool wear it causes, and stock form sets how much of what gets bought is actually removed as chips. Both are decided in the material call-out and the bar or tube size chosen, long before the drawing reaches a machine.
Free-machining grades such as 12L14 or C1117 cut faster and hold tool life longer than an alloy chosen mainly for hardness or fatigue strength, such as 52100 bearing steel. On the aluminum side, 6061-T6 machines predictably while 2011-T3 behaves closer to a true free-machining alloy; silicon bronze cuts cleanly but work-hardens if a dull edge rubs instead of cutting. None of that changes what a part does, so choosing the easier grade for a bracket or a pin, and reserving the harder one for a feature that actually needs its wear resistance or strength, usually costs nothing to do.
Stock form matters as much as the alloy. A part that is genuinely hollow through its length is a natural fit for ASTM A513 DOM tube; a part that only looks hollow on one end is usually cheaper made from a drilled bar, because qualifying a second raw-material stream for one feature costs more than the extra chips. The same logic applies to diameter: a size that falls between two stocked bar sizes forces either a special stock order or removing more material than the part needs just to reach a standard size, and either way that extra step shows up in the price.
Why do more setups, and a smaller order quantity, both raise the cost per part?
Every time a part is unclamped and re-oriented, the machine has to re-establish its reference, which costs time and sometimes a dedicated fixture. That cost is close to fixed per run, whether the run is for one part or many, so it lands harder on a small order.
A part machined complete in one setup pays that fixed cost once. A part that needs a second operation to reach a face or a bore on the far side pays it twice, and a fixture built to hold an odd shape for that second operation is itself a cost that has to be justified by the order behind it. Designing a part so its features are reachable from as few sides as possible is one of the few changes that lowers cost without touching a single dimension.
The same arithmetic governs first-article approval and any dedicated gauging: dialing in a program, cutting the first good part, and checking it against every called-out dimension takes roughly the same time regardless of batch size. On a very small run, that fixed time is divided across very few parts, so each one carries a larger share of it. Batching similar parts into fewer, larger releases spreads that same fixed cost over more pieces and brings the average down; it is one reason a supplier will ask about the realistic order range before quoting a firm number.
What makes cycle time longer inside the machine?
The features that take longest to cut are the ones that force a slower feed rate or an extra pass: numerous small features, pockets deep relative to their width, internal radii smaller than a standard cutter, and walls thin enough to deflect under cutting force.
A deep, narrow pocket cannot be cleared with a wide cutter, so it is roughed with a smaller tool taking lighter cuts and finished in more passes than a shallow pocket of the same footprint. An internal radius specified smaller than any standard end mill forces either a slower plunge-and-circle move with a reduced-diameter tool, or a dedicated one made for that single feature, instead of a clean pass with stock tooling already on the machine.
Thin walls add a different kind of time. They deflect away from the cutter under load, so the program has to cut lighter and sometimes add a spring pass to bring the wall back to size once it relaxes. Feature count matters just as much as any single hard feature: a part with many small chamfers, grooves and cross-holes spends real time indexing between them, even when none of those features is individually difficult.
How do tolerance and surface-finish call-outs raise the cost?
A tolerance narrower than the process needs slows the finishing pass and increases how often parts are checked, and a fine roughness call-out can add a pass or move a feature to a different process altogether. Both costs scale with how many features carry the tight number, not with the size of the part.
Turned diameters on the lead plant's CNC lathes are quoted against the drawing; the general tolerance classes set out in ISO 2768 cover the rest of a drawing without any individual call-out, as described on the CNC turning page and in the ISO 2768 and NADCA tolerance guide. A dimension inside the general class is unremarkable; a tolerance called tighter than the process needs, on a face that never mates with anything, adds cost without adding function.
Surface finish works the same way. A roughness call-out belongs on a seal, a bearing seat or a sliding face, the same distinction the surface finishing page draws for cosmetic-class call-outs and masking; a face nobody touches does not need one, and putting a call-out on every surface out of habit forces a slower pass or a cosmetic-inspection standard on features that will never be seen. When a coating is applied after machining, marking exactly which threads, bores or press-fit surfaces must stay bare turns masking into a single clean step instead of a round of questions before the part can run.
How much does inspection effort add to a CNC machining quote?
Inspection cost rises with how many features are checked, how tightly they are toleranced, and which instrument the tolerance forces the shop to use. A hand gauge on a loose dimension takes moments; a coordinate-measuring-machine report on a tight one takes much longer and is priced accordingly.
Every measurement carries its own uncertainty, and per the NIST guidelines on evaluating and expressing measurement uncertainty, a tolerance is only provably met once that uncertainty is taken into account. On a wide tolerance the uncertainty is a small share of the band and a hand tool is enough; on a tight one it can take up a large share, which is what pushes a feature toward a controlled-room CMM and a written measuring method instead of a quick check at the machine.
An RFQ that lists which features actually need a full dimensional report, as against which are covered by the general tolerance note, lets a quoting engineer price the inspection plan instead of assuming every dimension needs the same scrutiny. The machining RFQ guide lists what that package should include.
Why does a custom thread cost more than a standard gauge?
A standard thread checked with a go/no-go gauge the shop already owns adds almost nothing to the price. A custom pitch, form or class needs its own gauge set made or bought before the first part can even be checked, and that cost is carried entirely by the order behind it.
Straight and tapered pipe threads such as ANPT and NPT, gauged to ANSI/ASME B1.20.1, use go/no-go gauges the lead CNC plant already owns, because they are the pipe-thread forms it runs routinely. A thread specified to a non-standard pitch, a modified form, or a class outside that standard set has to be checked against a gauge made for that one job, and the cost of making or sourcing it does not disappear just because the run is small.
The same logic runs one level down, to knurling. Straight and diamond knurls are cut with a tool already set up on the machine, but every change of knurl pitch is a tool change, and a family of similar parts that all share one knurl pitch keeps running back to back instead of resetting between them.
Is it cheaper to cast a part and then machine it, or machine it from solid bar?
Casting trades a tooling cost and a longer time to first parts for a much shorter machining cycle once the tool exists, because most of the shape is already there before the part reaches a mill. Machining from solid skips the tooling cost entirely but pays, on every part, for every bit of material removed as chips.
Which route costs less depends on the shape, the quantity, and how much of a part's cost is stock removal versus assembly and finishing. The die-casting-versus-machining break-even guide works through that comparison in detail, including the point at which a mold's fixed cost stops mattering to the per-part price.
For a part that is a good casting candidate, the drawing discipline that lowers machining cost still applies once the casting reaches a second operation. A casting machined on the lead CNC plant's CNC milling centers still needs one consistent datum scheme shared between the casting drawing and the machining drawing, bore sizes rounded to standard tooling, and roughness call-outs left off any face that is never touched. None of that changes how the die casting looks off the tool, but each one changes how long the second operation takes.
None of the above is the landed cost of the part. Duty, tariff classification and processing fees such as the U.S. merchandise processing fee, assessed under 19 CFR 24.23, sit on top of a machining price and are covered separately in the tariffs and country-of-origin guide; they do not change why the part itself costs what it does.
Frequently asked
Can you just give me a typical price range for a part like mine?
Not usefully. Two parts that look similar in a photo can differ in machinability, setups, tolerance and quantity enough to land in very different places, so a range pulled from an unrelated part is more likely to mislead than help. The reliable number is a quote against the actual drawing.
Does a tighter tolerance always make a machined part better?
Only where the function needs it. A tolerance tighter than a mating, sealing or locating feature requires does not make the assembly work better; it adds inspection time and scrap risk on a dimension that was never going to be a problem in use.
Will ordering more parts always lower the price per piece?
Up to a point. A larger order spreads the fixed cost of setup, programming and gauging over more parts, which lowers the average. Past the point where that fixed cost is already a small share of the total, the price per part is set mostly by material and cycle time instead, and ordering still more does not keep lowering it at the same rate.
Should I use a standard thread even if a different one fits my design better?
Where function allows it, yes. A standard pipe thread gauged to ANSI/ASME B1.20.1 uses tooling and gauges a shop already owns; a thread outside that standard set is not wrong, but it is priced as its own gauge and inspection step.
How do I find out which of these is actually driving the cost on my part?
Send the drawing for a quote and ask which features or call-outs are driving the route, the setups, or the inspection plan. A useful quoting response names them feature by feature rather than returning a single number with no explanation.
Sources
- ASME B1.20.1-2013, Pipe Threads, General Purpose, Inch, ASME, 2013 (reaffirmed 2018)
- ASME Y14.5-2018, Dimensioning and Tolerancing, ASME, 2018
- Taylor and Kuyatt, NIST Technical Note 1297: Guidelines for Evaluating and Expressing the Uncertainty of NIST Measurement Results, NIST, 1994
- 19 CFR 24.23, Fees for processing merchandise, eCFR, accessed September 2026