Engineers tell me they want one prototype. One part. That's the answer I hear most days, and it's usually the wrong one.
A single machined part tells you the geometry is right. It does not tell you the design is right. Those are different things, and a lot of budget dies in that gap. This guide is about the quantity question — how many CNC prototypes to machine, what each quantity actually validates, and when more parts is just spending money on the same information twice.
I run a CNC shop. I have shipped thousands of prototype orders. The pattern is consistent: the projects that go smoothly ordered a number that matched the question they were trying to answer. The ones that stalled ordered "one" because nobody asked what the prototype was for.
What a prototype is actually for
A prototype is a test, not a deliverable. Before you pick a quantity, you need to know what you are testing. Four different questions, four different quantities:
| What you want to know | Minimum useful quantity | Why |
|---|---|---|
| Does the geometry match the drawing? | 1 | A single part confirms the model. Nothing about assembly or function. |
| Do the parts fit together? | 2-3 of each part | Fit is about relationships between parts. One part can't show you a fit. |
| Does the assembly work under load? | 3-5 of each | You will break some during testing. This is normal. |
| Is the process repeatable? | 5-10 | Repeatability is a statistical property. You cannot measure it with one sample. |
The expensive mistake is ordering the wrong quantity for the question. Machining ten parts when you only needed fit validation burns money. Machining one part when you needed repeatability sends you back for a second run — which costs more than the first run, because setup gets paid twice.
The four quantities that matter
I have seen prototype orders in every size. They cluster around four numbers, and each number maps to a specific goal.
One part: geometry check only
One prototype is for confirming the CAD model machines correctly — draft angles are feasible, pockets reach depth, threads tap clean. It is a manufacturing check, not a design check. If you order one part, treat it as a sample to hold in your hand and compare to the model, nothing more.
This is also the number to order when the cost of the part is dominated by setup, and you are confident in the design. The marginal cost of parts 2-3 is often small — you already paid for the program and the fixture. On aluminum parts, going from qty 1 to qty 3 typically adds 30-50% to the order, not 200%. The setup is the expensive part.
Two to three parts: fit and first assembly
The first thing people discover when they put machined parts together is that tolerances stack. A part at its high limit meets a mating part at its low limit, and suddenly the slip fit is an interference fit. Two or three of each part lets you check worst-case fit, not just the one lucky combination that happens to assemble.
Three copies is my default recommendation for the first prototype run of anything that will be assembled. It is cheap relative to what it protects you from: a mating design that only works by luck.
Three to five parts: functional testing
If you are going to load the parts, run the mechanism, cycle it a hundred times, you will break parts. Plan for it. Three to five copies means you can test to failure and still have a reference part left over to measure against.
This is the quantity for moving mechanisms, sliding fits, parts that carry load. You need enough samples that a single testing mistake does not end the experiment.
Five to ten parts: repeatability and production confidence
Repeatability is the thing you cannot see in one part. Machine five to ten identical parts and measure them all: the spread between the largest and smallest tells you whether this process holds tolerance run to run. If the spread is fine, your production order will be fine. If it is not, you found out on a prototype run instead of a 500-piece production order.
This quantity also gives you a realistic per-part cost for production. Setup gets amortized, and the per-part price you see at qty 10 is close to what you will pay at qty 500. That number is worth money when you are budgeting the real order.
How much this actually costs
Real numbers from our shop, a medium-complexity aluminum part (roughly 80mm x 60mm x 25mm, with pockets, holes, and threads):
| Quantity | Per-part price | Total order | What it validates |
|---|---|---|---|
| 1 | $145 | $145 | Geometry only |
| 3 | $95 | $285 | Fit and first assembly |
| 5 | $75 | $375 | Functional testing, some repeatability signal |
| 10 | $58 | $580 | Repeatability and a realistic production price |
The jump from qty 1 to qty 3 costs $140. The jump from qty 3 to qty 5 costs $90. Both are small compared to the cost of a failed production order, which is measured in weeks of lead time and scrap. The per-part price drops 60% between qty 1 and qty 10 — that is the setup cost being spread across more parts.
When a prototype is the wrong tool
Prototypes answer design questions. They are not the right tool for every question, and ordering more of them does not fix that.
Surface finish. If your question is "what will this look like anodized?" a machined prototype answers it partially. Anodize is affected by alloy, grain direction, and polishing — you may need to test finish on the actual production alloy and process, which is a different job than a geometry prototype.
Material behavior. If your question is "does this PEEK part hold up at 150C?" a prototype machined from different stock or a different grade tells you nothing. The answer depends on the exact grade and supplier. Match the material to the production spec before you draw conclusions.
Long-run durability. If your question is "will this fail after 10,000 cycles?" you need a fatigue test, not a prototype order. Prototype quantity is about design validation, not statistical reliability.
None of this means prototypes are useless. It means a prototype answers the question you designed it to answer, and you should pick the quantity that matches that question.
The setup trap
Here is the mistake I see most often from engineers who order aggressively: they machine one part, approve it, and then realize the design needs a change. The change is usually small — a hole moves 2mm, a wall thickens. But it is a change, which means a new program, a new setup, sometimes new tooling. They pay for the first setup and the second setup.
Three parts protects you here. If you machine three, you can test the assembly, find the problem before it is baked in, and fix the design in one revision instead of two. The cost of the extra two parts is usually less than the cost of a second setup, and it is certainly less than the cost of a production run based on an untested mating design.
This is why I tell engineers to order three unless they have a specific reason not to. Three is the default that matches the most common prototype questions, and it is cheap insurance against the most common prototype failure mode — approving geometry that does not actually assemble.
What to send so the quantity matters
The quantity question only helps if the parts are right. Two things make or break a prototype run more than anything else:
Send the STEP file and a drawing with tolerances you actually mean. The most common prototype delay is a drawing that says "aluminum" without a grade, or tolerances that are tighter than the function needs. A 6061 part and a 7075 part machine very differently, and a +-0.01mm tolerance on a feature that does not need it doubles inspection time. Be specific, and the quote and the lead time both come back better.
Tell the shop what the prototype is for. This sounds obvious, but it changes how a machinist works. If you tell me a bore is a bearing seat, I will check it differently than if it is a clearance hole. If you tell me the part carries load, I will flag a thin wall that might deflect. The shop can only protect you from problems they know about.
For the full file-preparation checklist, see our guide on how to prepare a CNC machining RFQ and the rapid CNC prototyping turnaround guide.
Matching quantity to your stage
A prototype order is usually one step in a longer sequence. The quantity that makes sense depends on where you are:
First iteration, uncertain design: order 3. You are going to change things anyway. Three gives you assembly feedback without committing to a quantity you will throw away.
Validated design, need production confidence: order 10. The goal is repeatability data and a real production price, not assembly feedback. Ten is the number that tells you what production will cost.
Bridge production, need parts now: skip prototyping, order production quantity. If you already have validated parts from a prior run, a prototype order is spending money on information you already have. Go straight to a production batch with small batch CNC machining.
The mistake is ordering a middle quantity for no reason. Ten parts when you only needed assembly feedback wastes money. Three parts when you needed repeatability data sends you back for a second run. Match the number to the question.
Ready to order the right number?
The quantity question has a real answer, and it is not "as few as possible." It depends on what you are trying to validate — geometry, fit, function, or repeatability. One part validates geometry. Three validates fit and assembly. Five gets you functional testing. Ten buys you repeatability and a realistic production price.
Send us your STEP file and drawing for a free DFM review and a prototype quote. We will tell you how many parts actually answer your question, and what they will cost — before you spend money on a quantity that does not. We machine CNC prototypes in aluminum, stainless, titanium, and engineering plastics, and we give every prototype order the same first-article inspection we give production runs.
Upload your files today. No commitment — just a straightforward answer to the quantity question and a quote you can budget against.