Samples of CNC Machining Services: What to Check Before You Commit
A sample is the cheapest place to find a defect. This guide is for design engineers and sourcing staff who need to judge a shop from the parts it sends back, not from a capability page. Read it and you will know which checks matter, which claims to ignore, and when a sample run is the wrong tool.

In this article
- 1
- 2
- 3
- 4
- 5
- 6
- 7
- 8
- 9
- 10
Key takeaways
Which sample route fits your part
Pick the row that matches your geometry, material and volume. The tolerance column is the tightest value we hold in production, not a one-off best result.
| Sample route | Best for | Tolerance / finish | Watch out for |
|---|---|---|---|
| 3-axis CNC | Prismatic parts, open faces | ±0.005 mm, Ra 1.6–3.2 μm | Undercuts need a second setup |
| 4-axis CNC | Shafts, housings with side holes | ±0.005 mm, Ra 0.8–1.6 μm | Rotary table limits part length |
| 5-axis CNC | Impellers, turbine blades, organic shapes | ±0.005 mm, Ra 0.8–1.6 μm | Programming time adds days |
| Mill-turn | Turned parts with milled flats | ±0.005 mm, Ra 0.8–1.6 μm | Not for very large diameters |
| CNC turning | Round parts, high volume | ±0.005 mm, Ra 1.6–3.2 μm | Off-axis holes need a mill |
| Die casting sample | Thin-wall housings, 1,000+ pcs | As-cast plus machining | Mold lead time is weeks |
| Vacuum casting | Small batches of plastic-like parts | Ra 1.6–3.2 μm | Not for structural loads |
Start with the drawing, not the sample
Samples of CNC machining services only mean something against a controlled drawing. Before you send a file, decide which dimensions actually matter. A print with 60 tight tolerances and no datum scheme forces the shop to guess, and guessing is where scrap starts.
Mark the critical-to-function dimensions. Those are the ones you will measure on the sample. Everything else can sit at general tolerance. This single step cuts quoting time and removes most arguments later.
Check that your 3D model and 2D drawing agree. When they disagree, machinists follow the drawing. If the model has the newer geometry, say so in writing.
- 1Define the datumOne primary datum, then secondary and tertiary. Without it, hole positions drift.
- 2Flag free-form surfacesCurved surfaces need either a scan report or a set of section points.
- 3State the material temper6061-T6 machines differently from 6061-O. Temper affects finish and thread strength.
Read the inspection report like an engineer
A sample without data is a paperweight. Ask for dimensional results from a coordinate measuring machine, tied to the serial number on the part. If the shop sends a generic report with no ID, the numbers prove nothing.
Look for the measurement method, not just the result. A Ø12 H7 bore checked with a plug gauge is fine. The same bore checked with calipers is not a real result at that tolerance.
Watch for out-of-tolerance values reported as "within drawing". If a dimension is 0.008 mm over and the drawing allows ±0.005 mm, it is out. Ask for the rework plan.
Material certs should match the heat number on the bar stock. For aerospace and medical work, traceability is part of the sample, not an extra.
- 1Serial number on the reportTies the data to the physical part in your hand.
- 2Method listed per dimensionCMM, micrometer, gauge, profilometer. Each has a range where it is valid.
- 3Material and finish certsAnodize thickness, plating thickness, hardness. Ask for them when the drawing calls them out.
Judge repeatability, not a single perfect part
A skilled machinist can hand-finish one part to spec. That tells you little about a 500-piece run. Ask for three samples from the same program and the same setup. Measure the spread.
If the three parts vary by 0.004 mm on a ±0.005 mm feature, the process is marginal. If they vary by 0.001 mm, the process is under control. This is the single most useful number in a sample review.
For plastic parts, repeatability also covers moisture and temperature. POM and PA absorb water, so dimensions move after machining. Let the sample sit in your facility for a day before final measurement.
- 1Same setup, three partsShows whether the process or the operator is holding the tolerance.
- 2Measure after stabilizationPlastics and thin-wall aluminum move after release from the vise.
- 3Note the ambient conditionsA 20 °C metrology room gives different numbers than a 30 °C shop floor.
Match certifications to your industry
Certificates are only useful when the scope covers your part. ISO 9001:2015 covers general quality management across the plant. IATF 16949:2016 is the automotive and EV standard, with stricter change control and traceability.
ISO 13485:2016 applies to medical devices and brings process validation and clean handling requirements. ISO 27001:2022 covers information security, which matters when you upload proprietary CAD files.
Ask to see the certificate number and expiry date. Then ask which processes are inside the scope. A shop can hold IATF 16949 for machining while sending plating to an uncertified subcontractor.
For medical and aerospace programs, ask how sub-tier suppliers are controlled. The answer tells you more than the certificate on the wall.
- 1ISO 9001:2015Baseline quality system. Covers most industrial and electronics work.
- 2IATF 16949:2016Automotive and EV. Requires PPAP-style documentation on request.
- 3ISO 13485:2016Medical devices. Adds validation and contamination control.
- 4ISO 27001:2022Information security for your files and drawings.
Ask what the sample does not prove
A machined sample does not prove die casting tool life, sheet metal fixture repeatability, or injection mold cycle time. Those processes need their own sampling step with the actual tool.
A sample also does not prove packaging, labeling, or shipping survival. If your part is fragile or cosmetic, request the shipping method before the run starts.
Surface finish on a sample can be hand-polished to look better than production. Ask whether the finish came from the cutting tool or from manual work. If it was manual, the production parts will not match.
- 1Tooling processes need a tooling sampleDie casting, injection molding and stamping prove out on the real mold or die.
- 2Cosmetic parts need a finish standardDefine the surface with a visual sample or a roughness value, not words like smooth.
- 3Packing is part of the sampleAsk for the actual box, tray or bag that will ship with the production order.
Compare lead time, MOQ and quote detail
A quote without a lead time is half a quote. Ask when the sample ships and when production can start after approval. At GreatLight, quotation and DFM analysis come back within 12 hours, production can start within 24 hours, and parts ship in 3–5 days.
MOQ is the other half. Some shops quote a low sample price, then require 500 pieces to start production. We run from one prototype to 10,000+ part runs with no minimum order quantity.
Read the quote line by line. Material, machining, finish, inspection and freight should be separate. If finish is bundled into machining, you cannot compare shops fairly.
For repeat orders, ask what triggers a price change. Material index, volume break and drawing revision should all be named in the quote.
- 1Lead time in writingSample ship date and production start date, not just a range.
- 2MOQ stated per processCNC and sheet metal differ from die casting and injection molding.
- 3Finish priced separatelyAnodize, plating and bead blasting each carry their own cost and lead time.
Protect your files before you upload
Your drawings contain the design intent. Ask how the shop stores them, who can open them, and how long they are kept. ISO 27001:2022 is the relevant standard here.
A mutual NDA is normal practice, not an insult. Send it before the RFQ, not after. We sign NDAs on request and treat uploads as secure and confidential.
Also ask whether the shop shows your parts in its public portfolio. Some do by default. If your program is under NDA, confirm that in writing.
- 1NDA before RFQCleaner than trying to retroactively restrict a drawing already shared.
- 2File retention policyAsk how long files stay on the server and how deletion is confirmed.
- 3No portfolio use without consentState it explicitly if the program is confidential.
How to run a sample order in 6 steps
This is the sequence we use with new programs. Each step has a gate, so a problem gets caught before the next cost is committed.
- 1Send the drawing and the modelInclude the datum scheme, critical dimensions, material temper and finish callout. Note any conflict between model and drawing.
- 2Request DFM feedback with the quoteLook for thin walls under 0.8 mm, deep pockets over 4× tool diameter, and tolerances tighter than the process holds. Ask for the quote and DFM analysis together.
- 3Approve the process planConfirm machine type, number of setups, stock size and inspection method. This is where 5-axis versus 3-axis gets decided.
- 4Order three samples, not oneSame program, same setup. Measure the spread on critical dimensions. Compare against the ±0.005 mm target.
- 5Inspect and sign offCheck the CMM report against the serial number, verify finish with a profilometer or visual standard, and confirm material certs.
- 6Release production with a frozen revisionLock the drawing revision and the process. Any later change restarts the sample gate, because geometry and tooling both move.
Questions buyers ask about samples
How many samples should I order?
Three is the useful minimum. One part shows what is possible, three parts show what is repeatable.
For cosmetic parts, add one extra to keep as a finish reference. Store it away from light and handle it with gloves.
Should I pay for samples?
Yes, in most cases. A free sample usually means the cost is hidden in the production price, which makes comparison harder.
What you should expect free is the DFM analysis and the quotation. Those are engineering time, not machine time.
Can a sample be made without a 3D model?
Yes, from a fully dimensioned 2D drawing. The shop will build the geometry from the print.
It adds programming time and raises the risk of misreading a dimension. A model plus a drawing is faster and cheaper.
What tolerance can a machined sample hold?
We hold ±0.005 mm (±0.0002 in) in production on critical features, with surface finish from Ra 0.2–0.8 μm up to Ra 1.6–3.2 μm as-machined.
Tighter than that needs a specific conversation about feature size, material and measurement method.
How fast can I get a sample?
Quotation and DFM analysis come back within 12 hours. Production can start within 24 hours of approval, and parts ship in 3–5 days.
Complex 5-axis geometry or special material orders add time. Ask for the schedule in writing before you commit.
Does a sample guarantee the production run?
No. It reduces risk, but production adds fixtures, tool wear and operator variation.
Ask how the first article is handled at production start. A first-article inspection on the real process closes that gap.
Send a drawing, get a quote and DFM notes
Upload your file and we will return a quotation with DFM analysis within 12 hours. No minimum order quantity, NDA on request.
12-hour quote±0.005 mm tolerance100% inspectionNDA on request