Two suppliers quoting the same drawing can return prices that differ by a large multiple, and neither is necessarily wrong. They have assumed different batch sizes, stock forms, tolerance readings and inspection scopes, and each assumption moves the answer. This article sets out the cnc part price factors as a structure rather than a price list: what each does to the cost, and which of them a buyer controls.
Key takeaways
- Every quote decomposes into three buckets: fixed per order, fixed per batch, and variable per part. Almost every surprise in a quote comes from the middle one.
- Setup amortisation is the single largest lever at low volume. The worked example below shows the per-part figure falling from about 21 times the marginal cost at one piece to about 1.2 times at a hundred.
- Geometry acts through tool size. A corner radius that forces a 4 mm cutter instead of a 12 mm one multiplies roughing time roughly eighteenfold in the worked example, for the same volume of metal.
- Tolerance costs twice — once in process and once in proof. Under ISO 14253-1:2017 the measurement uncertainty is subtracted from the tolerance zone before conformity can be declared.
- The buyer controls more than the supplier does. Batch size, tolerance marking, stock form and lead time are all decided before the RFQ is sent.
On this page
- The three buckets every quote falls into
- Why does the same part cost differently at different quantities?
- Material: three separate cost effects, not one
- How does geometry turn into machining time?
- Tolerance and inspection: paying twice
- Surface finish and secondary processes
- Lead time and scheduling
- What does the buyer actually control?
- When the lowest price is the wrong choice
- Frequently asked questions
The three buckets every quote falls into
Whatever format a quotation arrives in, the underlying structure is the same. Costs are fixed per order, fixed per batch, or variable per part, and knowing which bucket a line sits in tells you immediately whether quantity will help.
| Bucket | What sits in it | Behaviour with quantity | Who mainly controls it |
|---|---|---|---|
| Fixed per order | Quoting and engineering review, programming, documentation package, first article | Spread across the whole order; falls fast with quantity | Buyer, through drawing quality and repeat business |
| Fixed per batch | Fixturing, setup and alignment, proving the first part, tool preparation | Spread across the batch; the dominant term at low volume | Shared — batch size is the buyer’s, fixture design the supplier’s |
| Variable per part | Cycle time, material, tool wear, in-process inspection, finishing | Constant per part; dominant at high volume | Mostly design — geometry, tolerance, material, finish |
The table is diagnostic. If two quotes differ wildly at ten pieces and converge at a thousand, the difference is in the batch bucket — fixturing approach or setup count. If they differ at both, the difference is in how the drawing was read, usually tolerance or finish.
Why does the same part cost differently at different quantities?
Because the fixed buckets are divided by the number of parts, and at low quantity they are far larger than the part itself.
Worked example: how setup amortises
Let c be the marginal cost of producing one more part once the machine is running — material, cycle time, tooling, in-process checks. Let S be the fixed cost of the batch: fixture, setup, alignment and proving. Then the cost of each part in a batch of n is
cost per part = c + S / n
Assume, purely as an input to the arithmetic, that setting up this particular part costs the equivalent of twenty parts’ worth of running time, so S = 20c. Dividing through by c gives the multiple of marginal cost at each quantity:
| Batch quantity | Setup share per part (S/n) | Cost per part as a multiple of c | Reduction versus a single piece |
|---|---|---|---|
| 1 | 20c | 21.0× | — |
| 5 | 4c | 5.0× | 76 per cent |
| 10 | 2c | 3.0× | 86 per cent |
| 50 | 0.4c | 1.4× | 93 per cent |
| 100 | 0.2c | 1.2× | 94 per cent |
| 500 | 0.04c | 1.04× | 95 per cent |
Two conclusions follow. The steep part of the curve is over by about fifty pieces — beyond that, further quantity buys progressively less, and negotiating for a larger batch to chase a lower unit price stops paying. And the quantity at which setup falls below a tenth of the marginal cost is simply ten times the ratio S/c, which here is two hundred. If you know roughly how setup-heavy your part is, you know where its volume break sits without asking anyone.
This is also why splitting an annual requirement into many small releases is expensive: each release is a new batch and pays the batch bucket again. Where the schedule allows, machining once and releasing from stock is usually cheaper than machining four times. Larger repeat programmes are handled under CNC machining services, where the fixture is built once and reused.
Material: three separate cost effects, not one
Material is usually one line on a quote, but it acts in three independent ways, and confusing them leads to substitutions that save nothing.
| Effect | What it depends on | How to reduce it |
|---|---|---|
| Stock cost and form | Alloy, and whether the nearest standard bar or plate is close to the finished envelope | Design to a standard stock size; a 2 mm change in envelope can move the part to a smaller bar |
| Removal volume | How much of the billet becomes chips | Near-net stock, or a casting or extrusion blank at volume |
| Machinability | Cutting speed the material tolerates, and how fast it wears tools | Choose the least exotic alloy that meets the requirement |
The third effect is the one most often underestimated. Titanium alloys and nickel superalloys conduct heat poorly, so heat that should leave with the chip stays at the cutting edge; cutting speeds drop and tool consumption rises, both of which show up in cycle time rather than in the material line. Stainless grades work-harden, which punishes any dwell in the cut. Aluminium alloys cut freely, which is why on aluminium parts the setup bucket so often dominates the total. MW+ works across 70+ materials, and the substitution worth testing is rarely to a cheaper alloy — it is to a more machinable one of the same class.
How does geometry turn into machining time?
Mostly through tool size. Geometry sets the largest tool that can reach a feature, tool size sets the depth and width of cut and the feed the tool will tolerate, and those three together set the rate at which metal leaves the part.
Worked example: what an internal corner radius costs in roughing time
Material removal rate is Q = ap × ae × vf, where ap is axial depth of cut, ae is radial width of cut and vf is the feed rate. Roughing time for a volume V is simply V / Q.
Take a pocket measuring 100 × 60 × 20 mm, so V = 120,000 mm³, and compare two tools. The parameter values are illustrative inputs, not a claim about any specific machine.
Case 1 — a 12 mm cutter, permitted by a 6 mm internal corner radius: ap = 6 mm, ae = 6 mm, vf = 1,200 mm/min.
Q = 6 × 6 × 1,200 = 43,200 mm³/min → t = 120,000 / 43,200 = 2.8 minutes
Case 2 — a 4 mm cutter, forced by a 2 mm internal corner radius: ap = 2 mm, ae = 2 mm, vf = 600 mm/min.
Q = 2 × 2 × 600 = 2,400 mm³/min → t = 120,000 / 2,400 = 50 minutes
Same pocket, same material, roughly eighteen times the roughing time — caused by a corner radius that is very often there for no functional reason. In practice a shop would rough with the large tool and clean the corners with the small one, which recovers most of the difference; but that is a second operation with its own tool change, and it only works if the pocket is open enough for the large tool to enter at all.
| Geometry feature | Mechanism | Design change that reduces it |
|---|---|---|
| Small internal corner radii | Caps the tool diameter for the whole pocket | Increase the radius; it is free to change in CAD |
| Deep pockets and bores | Long tools deflect, forcing lighter cuts | Open the cavity, reduce depth, or split the part |
| Features on many faces | Each face may need a setup, paying the batch bucket again | Consolidate features, or accept multi-axis machining to cut setups |
| Thin walls and unsupported spans | Part deflects under cutting force; feeds must drop | Thicken, add a rib removed last, or support in the fixture |
| Non-standard hole sizes | Forces interpolation or a special tool instead of a stock drill | Round to standard drill diameters wherever fit allows |
The article on when complex geometry requires multi-axis machining takes the setup-count question further, including the tolerance that each refixture consumes.
Tolerance and inspection: paying twice
A tight tolerance is charged once in the process that produces it and again in the measurement that proves it, and the second charge is the one that surprises people.
On the process side, a tighter band means lighter finishing passes, more frequent tool changes or offset corrections, thermal settling time and sometimes a different process route altogether. On the proof side, ISO 14253-1:2017 requires the measurement uncertainty to be subtracted from the specification zone before conformity can be declared — so a tighter band both needs a better instrument and leaves less of itself available to production.
| Tolerance decision | Effect on the process | Effect on inspection |
|---|---|---|
| Everything left to the general tolerance note | Standard passes, standard tooling | Sample check with hand instruments |
| A few features marked critical | Separate finishing operation on those features only | Targeted CMM routine on the marked features |
| Whole drawing tightened | Every feature becomes a finishing operation | Full CMM routine on every part, with the instrument qualified under ISO 10360-2:2009 |
| Geometric control added where function needs it, under ISO 1101:2017 or ASME Y14.5-2018 | Often no change — the feature was already being cut this way | A specific measurement, usually cheaper than tightening size |
The middle row is where most drawings should sit, and getting there costs nothing but a review. MW+ works to ±0.01 mm against ISO 2768-m generally and down to ±0.001 mm where a feature genuinely requires it, with Cpk ≥1.67 on controlled characteristics and a CMM report on every order. The companion article on tolerance standards for precision machined parts covers how to express the distinction on the drawing, and CNC machining quality control sets out the evidence that comes back.
Surface finish and secondary processes
Finishing is a separate cost structure with its own fixed and variable terms, and it is often quoted as a flat adder when it is really neither flat nor small.
Three things drive it. The parameter and value decide whether the finish comes off the cutter or needs a further operation; texture is defined under ISO 21920-2:2021 (superseding ISO 4287), with sampling and filtering under ISO 21920-3:2021, so a bare roughness number without a sampling length is ambiguous. The coverage matters: a finish called out on the whole part rather than on the faces that need it multiplies handling. And masking is labour: every surface that must be excluded from a coating is manual work on every part, and it does not amortise.
MW+ produces Ra 3.2 µm as-machined, Ra 0.4 µm fine-machined and Ra 0.1 µm polished. The practical rule is to specify the functional requirement on the functional surfaces and leave the rest as-machined; cosmetic requirements applied globally are one of the most reliable ways to inflate a quote without improving the part.
Lead time and scheduling
Lead time affects price through scheduling rather than work content. A part that can wait is fitted around existing setups and run when the right machine is free; a part that cannot wait displaces something, and the displacement is what is paid for.
The useful distinction is between a deadline and a preference. A genuine date — a qualification build, a line stoppage — is worth paying for. A default date entered on a purchase requisition is not, and relaxing it is one of the few concessions a buyer can make that costs nothing. MW+ quotes within 24 hours, and runs standard prototypes in 3–5 business days, 48-hour express where the programme needs it, and volume production in 10–15 business days with no minimum order quantity. Where geometry is still being proven, CNC prototyping first is usually cheaper than committing a production fixture to a design that will change.
What does the buyer actually control?
More than is generally assumed, and almost all of it before the RFQ is sent.
| Decision | Which bucket it moves | Typical leverage |
|---|---|---|
| Batch size and release schedule | Fixed per batch | Very high at low volume; negligible above the volume break |
| Marking which tolerances are critical | Variable per part, and inspection | High on drawings where tolerances were inherited from a template |
| Internal corner radii and pocket depths | Variable per part | High; often the largest single change available |
| Material class and stock form | All three | Moderate to high, and frequently overlooked |
| Finish coverage and masking | Variable per part | Moderate, and entirely within the drawing’s gift |
| Lead time | Scheduling | Moderate; free to give away when the date is nominal |
| Completeness of the RFQ package | Fixed per order | Moderate, plus a large reduction in revision cycles |
A complete package is a solid model, a drawing that marks the critical features, the material and any required certification, the finish with its parameter and coverage, the quantity and the expected repeat pattern, and the inspection evidence required. MW+ quotes from STEP, IGES, DXF, DWG, SolidWorks and PDF; you can request a CNC machining quote with those attached, and the guide on requesting an accurate quote lists the package in detail. The related analysis of what drives machine shop rates covers the supplier’s side of the same equation.
When the lowest price is the wrong choice
Four situations make the cheapest quote the most expensive outcome.
The quote assumed a looser drawing than you meant
A quote that is far below the others usually reflects a different reading of the tolerance or finish, not a better process. The parts arrive conforming to the supplier’s reading and failing yours. The fix is to confirm the interpretation before placing, not to assume the low bidder found an efficiency.
The documentation is not included
A certificate of conformance, a dimensional report and material certificates take real time to produce. Where a quote omits them and the programme needs them, the cost reappears later. Every MW+ order ships with a certificate of conformance, a CMM inspection report and material certificates under ISO 9001:2015, AS9100D, ISO 13485, IATF 16949 and NADCAP approvals; comparing quotes means comparing the same deliverable.
The volume break was bought with inventory
A lower unit price at a larger quantity is only a saving if the parts are used. If the design is still moving, or demand is uncertain, the amortisation curve above stops being a discount and becomes obsolete stock. The curve flattens around fifty pieces anyway, so the incremental gain from a very large commitment is small.
Single-sourcing a critical part on price alone
Where a part is on a critical path, the cost of a supply interruption dwarfs the unit price difference. Capability, capacity and documented process control are the relevant criteria, and CNC precision parts for regulated programmes should be qualified on evidence rather than on quotation.
Frequently asked questions
Why do two suppliers quote the same drawing so differently?
Almost always because they made different assumptions, not because one is more efficient. The usual candidates are the number of setups assumed, whether a dedicated fixture is being built, how the general tolerance note was read, whether the finish applies to the whole part, and what inspection evidence is included. Asking each supplier to state those five assumptions turns two incomparable numbers into a real comparison.
Is it cheaper to order a year’s requirement at once?
Up to a point. As the worked example shows, the steep part of the amortisation curve is over by roughly fifty pieces, so the gain from committing to a year rather than a quarter is usually small — and it is offset by carrying cost and by the risk that the design changes. A better structure is often one setup with scheduled releases from finished stock, which captures the batch saving without the commitment.
Does a cheaper material always make a cheaper part?
No. Material is usually a minority of the total on a machined part, and machinability affects cycle time, which is usually a larger share. Substituting to a cheaper but less machinable alloy frequently raises the total. The substitutions that do pay are to a more machinable grade in the same family, or to a stock form closer to the finished envelope.
How much does tolerance really move the price?
It depends on where the tolerance sits relative to what the process naturally holds. Tightening a band that the process already achieves comfortably changes almost nothing except inspection. Tightening past that point changes the process route, and the step is not gradual — it is the difference between one finishing pass and a separate operation with its own setup, measurement and settling time. This is why identifying the few features that matter beats adjusting a general tolerance class.
Do prototypes cost more per part than production?
Substantially, for structural reasons rather than pricing policy: a prototype carries the whole fixed-per-order and fixed-per-batch cost on one or a few pieces. That is the cost of proving the design, and far cheaper than finding the same problem after a production fixture has been built.
Can I reduce cost without changing the design?
Yes, and it is usually worth doing first. Confirming the general tolerance interpretation, restricting the finish to the faces that need it, relaxing a nominal delivery date, consolidating releases into fewer batches and supplying a complete RFQ package all reduce cost without touching a dimension. Design changes come next, and corner radii are the usual place to start.
What should I ask a supplier to break out on a quotation?
The split between the three buckets — what is charged once per order, once per batch, and per part — plus the assumed number of setups and the inspection scope. That breakdown is what lets you see whether a larger batch, a design change or a tolerance review is the lever that helps. A single figure with no structure cannot be acted on, however competitive it looks.
