A prototype and a production part can come off the same machine, in the same material, against the same drawing, and still be priced an order of magnitude apart. Nothing about the part changed. What changed is how many times the fixed cost of getting ready to cut gets divided. Reading a quote for prototype vs mass production CNC work is mostly reading that one division correctly.
This page sets out the cost structure behind the curve, the arithmetic that produces its shape, and the point at which ordering more stops rewarding you. It quotes no prices deliberately: unit cost depends on geometry, material, tolerance and finish, so any figure printed on a web page would be wrong for your part. What is transferable is the shape of the relationship.
Key takeaways
- Unit cost follows C(n) = v + S/n, with v the variable cost of one part and S the fixed cost of the job. That hyperbola falls steeply at low quantity, then flattens.
- A usable rule drops out of the same equation. Unit cost is within 10% of its floor once quantity reaches about 10 × (S/v). Below that you are still buying setup, not parts.
- Specification does not change across the curve. MW+ holds ±0.01 mm general to ISO 2768-m, ±0.005 mm precision and ±0.001 mm on selected features, on a one-off and a 100,000-piece release alike.
- What changes is process: dedicated fixturing, proven toolpaths, and sampling under a capability model such as ISO 22514-2 replacing 100% measurement.
- Documentation escalates with the programme, not the price. Every order ships with a certificate of conformance, a CMM inspection report and material certificates; AS9102 first article and PPAP Level 3 are quoted per programme.
- Scaling is the wrong decision while the design is unfrozen: a change after fixtures, gauges and an approved submission exist is the most expensive event on the curve.
- What actually changes between a prototype and a production run?
- Why does the per-part price fall as quantity rises?
- What changes at each quantity band
- Does the quality system change between prototype and production?
- How does material purchasing change with volume?
- Design freeze is the variable that dominates the curve
- When scaling up is the wrong call
- How MW+ runs the prototype-to-production transition
- Frequently asked questions
- What to send for a comparative quote
What actually changes between a prototype and a production run?
Almost nothing about the part. The drawing, the material and often the machine are the same. What changes is the plan around the cut: how the part is held, how often a human touches it, how the toolpath was arrived at, and how many pieces the preparation is spread across. Every cost element behaves in one of three ways as quantity rises, and sorting your part into those buckets is the whole of the analysis.
Fixed cost: paid once, whatever the quantity
CAM programming, collision simulation, fixture design and manufacture, machine setup, and proving the first piece against the drawing. None of it scales with quantity: a five-piece order and a five-thousand-piece order carry an almost identical fixed block. That is the entire reason low-quantity parts look expensive.
Variable cost: paid again for every part
Spindle time, material consumed, load and unload labour, deburring, and the fraction of an insert or end mill worn away per piece. This is the floor the curve approaches; no volume commitment takes you below it.
The semi-variable middle, where most surprises live
Inspection, finishing and logistics sit between the two. Inspection starts as a full dimensional report on every piece and becomes a sampling plan once the process has demonstrated capability. Finishing carries a batch element plus a per-piece element, so it behaves like a second, smaller cost curve bolted onto the first.
| Cost element | Behaviour as quantity rises | Bucket |
|---|---|---|
| CAM programming and simulation | Unchanged; divided across more pieces | Fixed |
| Fixture design and manufacture | Unchanged, but justifies a better fixture at volume | Fixed |
| Machine setup and first-off proving | Unchanged per setup; fewer setups per piece at volume | Fixed |
| Spindle and cycle time | Falls per piece as the programme is optimised, then constant | Variable |
| Material consumed | Constant per piece; utilisation improves with near-net stock | Variable |
| Dimensional inspection | 100% early, sampled once capability is demonstrated | Semi-variable |
| Surface finishing | Batch element plus per-piece element | Semi-variable |
| Programme documentation (FAI, PPAP) | Paid once per part number, per customer requirement | Fixed |
Why does the per-part price fall as quantity rises?
Because a constant is being divided by a growing number. Write the fixed block as S and the variable cost of one part as v. The cost of a batch of n parts is S + nv, so the unit cost is:
C(n) = v + S/n
That is a hyperbola, with a horizontal asymptote at C = v. Everything buyers call “the cost curve” is this one equation, and most intuitions about volume pricing are either a restatement of it or a mistake about it.
Worked example: amortising a fixed block
Take a mid-complexity milled part whose fixed block is fifty times the variable cost of one piece, so S = 50v. Results are expressed as multiples of v, so no currency is involved.
| Quantity n | Setup share S/n | Unit cost C(n) = v + S/n | Against the 5-off unit cost | Setup as % of unit cost |
|---|---|---|---|---|
| 1 | 50.00 v | 51.00 v | +364% | 98% |
| 5 | 10.00 v | 11.00 v | baseline | 91% |
| 25 | 2.00 v | 3.00 v | −73% | 67% |
| 50 | 1.00 v | 2.00 v | −82% | 50% |
| 250 | 0.20 v | 1.20 v | −89% | 17% |
| 500 | 0.10 v | 1.10 v | −90% | 9% |
| 5,000 | 0.01 v | 1.01 v | −91% | 1% |
Read the last two rows together. Going from 5 pieces to 500 removes 90% of the unit cost. Going from 500 to 5,000 — a tenfold increase in commitment, inventory and risk — removes a further 0.09 v, under 8% of the 500-piece price. The curve has already given you almost everything it has, which is the most useful thing to know before trading an annual volume for a price break.
The 10 × (S/v) rule
Ask when unit cost is within 10% of its floor. That needs S/n ≤ 0.1v, which rearranges to:
n ≥ 10 × (S / v)
With S = 50v that gives n ≥ 500, matching the table. S/v is the only quantity you need to estimate, and a supplier can give it without disclosing rates: ask what proportion of the five-off unit price is setup. If the answer is 90%, then for that five-piece batch S = 45v and the flattening point is near 450 pieces. A simple part with quick fixturing flattens early; a five-axis part needing a bespoke fixture keeps improving well past the quantity a buyer expects.
What changes at each quantity band
The equation explains price but not process. A shop changes how it makes the part at recognisable thresholds, and those changes move v downward as well as spreading S.
| Band | Workholding | Toolpath | Inspection regime | Typical MW+ schedule |
|---|---|---|---|---|
| 1–10 pieces | Standard vice, soft jaws, tombstone | Conservative, proven on the machine | Full dimensional report on each piece | 48-hour express or 3–5 business days |
| 11–100 pieces | Soft jaws cut to the part, simple stops | Optimised after the first pieces run | Full report on first article, then reduced | 3–5 business days to short-run schedule |
| 101–1,000 pieces | Dedicated fixture, multiple parts per setup | Tool life characterised, feeds raised | Sampling plan plus in-process gauging | 10–15 business days |
| 1,000–10,000 pieces | Dedicated fixture plus automation or bar feed | Cycle time engineered, offsets automated | Statistical sampling, capability tracked | 10–15 business days per release |
| 10,000+ pieces | Cell dedicated to the part family | Balanced across machines | Capability-based sampling, documented control plan | Scheduled against the release programme |
The step from the second band to the third costs the most and saves the most. A dedicated fixture is a capital item, and it is what allows several parts per cycle and loading without re-indicating. If lifetime volume will never reach the third band, that fixture is waste — the argument for staying in low-volume machining and trading cost against speed.
Does the quality system change between prototype and production?
The specification does not; the verification method does. A prototype is measured exhaustively because there is no process history. A production part is measured to a sampling plan because the process has demonstrated it holds the characteristic. Both are anchored in published standards rather than shop custom.
| Standard | What it governs | Prototype stage | Production stage |
|---|---|---|---|
| ISO 2768-1 | General tolerances for linear and angular dimensions | Governs everything not explicitly toleranced | Unchanged |
| ISO 286-1 / ISO 286-2 | ISO code system for tolerances on linear sizes, fits such as H7 | Applied to bores and shafts as drawn | Unchanged; gauging often added |
| ISO 1101 and ASME Y14.5 | Geometrical tolerancing and datum reference frames | Datums drive fixture design | Datums drive gauge design |
| ISO 10360-2 | Acceptance and reverification of CMMs for linear dimensions | Defines the measuring instrument’s permissible error | Unchanged |
| ISO 22514-2 | Process capability and performance models | Not yet applicable — no process history | Basis for sampling and for a Cpk ≥1.67 target |
| AS9102 | First article inspection reporting for aerospace | Performed at the transition, not on every trial piece | Re-run on defined changes |
| IATF 16949 | Automotive quality management, including PPAP submissions | Design records assembled | PPAP Level 3 submitted before release |
| ISO 9001:2015, AS9100D, ISO 13485 | Quality management system requirements by sector | Same system applies to both | Same system applies to both |
Because the specification is constant, a part validated as a prototype needs no re-qualification when it scales, provided the same shop and process route are used. That continuity is the main structural argument against splitting prototype and production between two suppliers. The metrology behind it is set out under CNC machining quality control, and the measurement sequence in our guide to the CMM inspection process.
How does material purchasing change with volume?
A prototype is cut from whatever stock is on the rack, because waiting for a mill delivery defeats the purpose. That removes more material than the part requires, costing stock and cycle time. At volume the stock size is specified close to the finished envelope.
Worked example: material utilisation
Take a part with an envelope of 60 × 40 × 20 mm, so a finished volume of at most 48,000 mm³. From a shelf offcut of 70 × 50 × 25 mm the stock volume is 87,500 mm³, so utilisation is 48,000 ÷ 87,500 = 54.9%. From near-net 63 × 43 × 22 mm bar ordered for a production release, stock volume is 59,598 mm³ and utilisation rises to 48,000 ÷ 59,598 = 80.5%.
That is a 31.9% cut in stock per part, and roughing volume falls from 39,500 mm³ to 11,598 mm³ — 70.6% less material turned into chips. In a hard-to-machine alloy the cycle-time effect of the second figure outweighs the stock saving.
| Decision factor | Points to the prototype band | Points to a production release |
|---|---|---|
| Design maturity | Revisions still expected after testing | Drawing frozen under change control |
| Lifetime volume | Below roughly 10 × (S/v) pieces | Comfortably above that threshold |
| Schedule pressure | Parts needed inside a week | Release dates can be planned weeks ahead |
| Verification | Piece-by-piece dimensional report wanted | Capability-based sampling acceptable |
| Fixturing | Soft jaws and standard workholding suffice | Dedicated fixture can be amortised |
| Documentation | COC, CMM report and material certificates | AS9102 first article or PPAP Level 3 required |
Design freeze is the variable that dominates the curve
Every figure above assumes the drawing stays still. An engineering change during prototyping consumes a re-program, a new billet and a repeat first-off check. The same change after a production release can invalidate a dedicated fixture, a set of gauges, an approved first article or PPAP submission, and finished stock.
So the right question is rarely “how many should I order for the best price?” but “how confident am I that this drawing is final?” Where confidence is low, stay in the prototype band deliberately and pay the setup premium as the price of optionality — which is what CNC prototyping exists to buy. Where confidence is high, moving up a band early captures the steep part of the curve.
Signals that a design is genuinely frozen
- Functional testing is complete on parts made by the intended process, not a substitute.
- Every critical dimension has a tolerance and a datum scheme per ISO 1101 or ASME Y14.5.
- Material and finish are specified to a standard rather than by trade name, and mating parts have been measured rather than only modelled.
- A change-control route exists, so a future revision is a decision rather than an email.
When scaling up is the wrong call
The curve rewards volume, so the default advice is to order more. It is wrong in at least four common situations.
Lifetime volume sits below the flattening point
If you will never buy 10 × (S/v) pieces in the product’s life, dedicated fixturing and process engineering will not be recovered. Stay in the band you are in and put the money into design.
Demand is uncertain, seasonal, or regulated
A lower unit price bought with unsold inventory is not a saving. Carrying cost, obsolescence risk and tied-up working capital all sit outside the quote. In regulated markets a pending approval can turn a full warehouse into scrap.
The part should not be machined at all at that quantity
Beyond a certain volume and geometry, casting, metal injection moulding, cold forming or stamping make the part more cheaply, with machining reserved for critical features. A supplier who quotes ever-larger machined quantities without raising this is optimising their own utilisation, not your cost.
Speed is worth more than unit cost right now
During validation a week of schedule is often worth more than the whole unit-cost difference. That is the argument behind a 3–5 day rapid prototype lead time: you are buying calendar, not parts.
How MW+ runs the prototype-to-production transition
MW+ has machined precision components in Shenzhen since 2015 — more than ten years — from a 15,000 m² facility running 60+ CNC machining centres with 120+ engineering and quality professionals. Over a million parts have shipped to 50+ countries against a 99% on-time delivery record, with 70+ materials in regular use.
The point for anyone reading a cost curve is that both ends run on the same floor: the programme that proved the prototype is the one optimised for the release. Prototypes ship in 3–5 business days, or 48 hours on the express route; volume production runs to a 10–15 business day schedule with no minimum order quantity and programmes past 1,000,000 units. Quotations return within 24 hours from STEP, IGES, DXF, DWG, SolidWorks or PDF.
Work runs under ISO 9001:2015, AS9100D, ISO 13485, IATF 16949 and NADCAP accreditation, with ITAR, RoHS and REACH registration. Every order ships with a certificate of conformance, a CMM inspection report and material certificates; AS9102 first article inspection and PPAP Level 3 are quoted per programme. Prototype work starts at CNC prototyping, scale-up is handled through machine parts manufacturing, and the full process list sits under CNC machining services. Finished families are catalogued as precision machined products, and the mechanics of moving a single part number up through the bands are covered in scaling CNC production from 1 to 100,000 parts.
Frequently asked questions
At what quantity does production pricing beat prototype pricing?
There is no universal number: it is set by S/v for your part. Ask what proportion of the five-off unit price is setup, then apply n ≥ 10 × (S/v). A simple turned part with a short program may flatten under a hundred pieces; a five-axis housing with a bespoke fixture may keep improving past a thousand. A supplier quoting one industry-wide crossover quantity is guessing.
Can I ship prototype parts to a customer as production parts?
Technically yes, if they conform and the inspection record proves it. Commercially, check what your quality system requires. A prototype batch carries a certificate of conformance, a dimensional report and material certificates, but no PPAP submission, control plan or capability index. If the purchase order calls for those, prototype parts do not satisfy it however accurate they are.
Will the tolerance get worse when my part goes to volume?
It should not, and at MW+ it does not: the same ±0.01 mm general band to ISO 2768-m, ±0.005 mm precision band and ±0.001 mm on selected features apply at every quantity. What changes is how conformity is demonstrated. On a prototype you get a measurement of the piece in your hand; in production you get evidence that the process holds the characteristic, as a capability index against ISO 22514-2 with a Cpk ≥1.67 target.
Is committing to an annual volume worth the discount?
Compare the discount against the flat part of the curve first. If your quantity already sits past 10 × (S/v), the remaining setup share is under 10% of unit cost, so no commitment can move the price much and you trade real inventory risk for a small number. Commitments pay when they cross a band threshold, not when they move you along one you are already inside.
Why is my second order not much cheaper than the first?
Programming is reusable, so that part of S does not recur. Setup is paid every time the job goes back on a machine, and fixtures must be refitted and re-indicated. A repeat order at the same quantity removes the programming share but not the setup share. Say so at the first RFQ if repeat orders are foreseeable — a fixture designed to be retained changes the arithmetic on every release.
Does splitting prototype and production between two suppliers save money?
On the quote, sometimes. On the programme, usually not. The second supplier pays the fixed block again — new programming, new fixturing, new first article — and the process that validated your design is not the one making your parts. Re-qualification and a different tool path producing a different residual stress state are real costs that appear on neither quotation.
How should I present quantities in an RFQ to see the curve?
Ask for a price-break table at several quantities in one RFQ — 5, 25, 100, 500 and 2,000, say — rather than quoting one quantity and negotiating later. The shape of the returned table gives you the supplier’s S/v directly, and it costs the estimator little: the fixed block is calculated once.
What to send for a comparative quote
To get a curve rather than a single number, send a 3D model in STEP or IGES with a dimensioned 2D drawing carrying the datum scheme, the material specified to a standard rather than a trade name, the surface finish requirement, and the quantities you want priced. State which dimensions are critical and which fall under the general tolerance note; that distinction changes the inspection plan and the fixture design, and therefore S. Say where you are in the design cycle — if the drawing is still moving, a supplier should steer you toward the prototype band, not a volume commitment.
