Testing a CNC Machining Service: 7 Checks Before You Send a PO
How engineers and buyers qualify a new machine shop without wasting a production cycle. This guide covers the proof you can ask for, the numbers you can verify, and the points where most suppliers quietly fall apart. Read it before you release a drawing.
In this article
- 1
- 2
- 3
- 4
- 5
- 6
- 7
Key takeaways
What to verify when testing a CNC machining service
Each row is one claim a supplier makes, and the evidence that either backs it or does not.
| Claim to test | Acceptable evidence | Red flag |
|---|---|---|
| Tolerance capability | CMM report on your part, measured values listed | "We hold ±0.01 mm" with no report |
| Inspection coverage | In-process and final inspection records | Only a packing slip in the box |
| Certification scope | Certificate number and issuing body, current | A logo on the website only |
| Process chain | Machining and finishing under one roof | Finishing farmed out to a third party |
| Lead time | Written dates for sample, then batch | Verbal "about a week" answers |
| Order size | One-off and 10,000+ runs both quoted | Minimum order not stated until PO stage |
| Data handling | NDA available before drawings are sent | Drawings requested by personal email |
Test the process, not the pitch
Pick the supplier who shows measured data on your part, keeps machining and finishing under one roof, and answers a technical question with a number. Everything else is a brochure.
Testing CNC Machining Service Tolerance Claims Against Real Measurements
Every shop quotes a tolerance number. Fewer can show you where that number came from. When testing a CNC machining service, the first thing to establish is whether the quoted tolerance applies to your part geometry or to a favorable demo part. A 20 mm aluminum bracket and a 900 mm steel frame are not the same problem, even if both quotes say ±0.005 mm.
Ask for the measurement method, not just the result. A caliper reading at ±0.02 mm resolution cannot verify a ±0.005 mm callout. A coordinate measuring machine can, and so can a micrometer on a simple diameter. If the shop cannot name the instrument used on each critical dimension, the tolerance claim is marketing.
The useful question is repeatability. One good part proves nothing. Ask how many pieces were measured in the last run of a similar part, and what the spread was. A shop that tracks spread knows its process. A shop that only reports pass or fail is guessing on your behalf.
Finally, match the tolerance to the feature. Bore diameters, flatness on a sealing face, and position of a bolt pattern each fail differently. Send a drawing with the critical dimensions marked and see whether the response addresses them individually or repeats one blanket number.
- 1Instrument matters±0.005 mm needs a CMM, micrometer or gauge, not a caliper.
- 2Ask for spread, not a single readingRepeatability across a run is the real capability.
- 3Mark critical dimensionsSee if the answer treats them separately.
Inspection Reports, Datum Schemes and What a Quote Should List
A quote is a technical document, not a price tag. When you compare two suppliers, the lower number often comes from a thinner scope: no deburring, no surface finish callout, no inspection report, no packaging spec. Those omissions reappear later as rework or as a rejected lot at incoming inspection.
A quote worth comparing lists the material grade, the machining processes, the finish with an Ra range, the inspection level, and the packaging. On a part with a sealing face, the finish matters as much as the dimension. Ra 0.8–1.6 μm is a normal machined finish; Ra 0.2–0.8 μm needs a deliberate step and shows up in the price.
Datum schemes are the quiet failure point. If your drawing calls out position relative to A-B-C and the shop machines from a single vise stop, the parts may measure fine individually and still not assemble. Ask how the setup references the datums, and whether a single setup covers the critical faces.
For anything regulated, request the inspection documentation up front. A first article report, material certificates and a final inspection record are normal deliverables. A supplier who treats them as extras will treat your deadlines the same way.
- 1Finish callouts change priceRa 0.2–0.8 μm is not the same quote as Ra 1.6–3.2 μm.
- 2Datum-driven setupsIndividual parts can pass and the assembly still fail.
- 3Documentation is a deliverableFirst article, material certs, final inspection.
Certifications, Process Chain and Lead Time Claims
Certifications are a filter, not a score. ISO 9001:2015 tells you the shop runs a documented quality system. IATF 16949:2016 matters if you build automotive parts. ISO 13485:2016 matters for medical devices. ISO 27001:2022 speaks to how your drawings and CAD files are handled. Check the certificate number and the scope, not just the logo.
The process chain is the second filter. A part that is machined in one place, anodized in another, and laser marked in a third has three schedules and three chances to lose a tray. Full-process capability under one roof shortens the path and keeps one team accountable when a finish drifts out of spec.
Lead time claims deserve the same skepticism as tolerance claims. Separate the sample date from the batch date, and ask which steps drive each. Cutting time is rarely the bottleneck; finishing, heat treat and inspection queues usually are. A shop that quotes 3–5 days for parts should be able to say which step is on the critical path.
Late delivery risk is worth asking about directly. Ask how many jobs slipped last quarter and why. No shop has a perfect record. A shop that answers with a specific cause is easier to plan around than one that answers with a percentage and no detail.
- 1Match the certificate to the industryIATF 16949 for automotive, ISO 13485 for medical.
- 2Count the handoffsEvery outside step adds a schedule and a risk.
- 3Ask which step is criticalIt is usually finishing or inspection, not cutting.
Order Size, Data Handling and the Cost of Switching
A partner who will run one prototype and a partner who wants 5,000 pieces are different businesses. If your program moves from prototype to pilot to production, you want the same shop to follow you through all three, or you will pay the learning curve twice. Ask what the smallest and largest runs of the past year looked like.
Setup cost dominates small runs, and per-piece cost dominates large ones. That is why a supplier refusing small orders is telling you something about its business model, not about your part. It is also why a quote for 1 piece and a quote for 1,000 pieces should come from the same process plan, not from two different shops.
Data handling is a real engineering concern, not paperwork. Your drawings carry geometry, tolerances and sometimes a customer's product architecture. Ask where files are stored, who can open them, and whether an NDA can be signed before you upload. A supplier who treats this lightly is a liability on any program with IP in it.
The last cost is switching. If the first shop fails, how fast can a second one pick up the process? Keep the drawing package clean: one 3D model, one 2D drawing with datums and finishes, one material spec. Clean inputs shorten the next quote from days to hours.
- 1One shop across the programPrototype, pilot and production on the same process plan.
- 2NDA before uploadNot after the drawing is already in someone's inbox.
- 3Clean drawing packageModel, drawing with datums, material spec.
Step by step: testing a CNC machining service in one week
A short paid trial that produces evidence you can compare across two or three suppliers.
- 1Send a drawing with critical features markedPick one part with a real tolerance, a finish callout and at least one datum-driven feature. Ask for a quote and a DFM note within a day.
- 2Compare the quote's technical scopeLook for material grade, Ra range, inspection level and packaging. If a column is missing, ask for it in writing before comparing price.
- 3Order one paid piece, not a free sampleA paid part follows the real production route. Free samples often come from a different setup and tell you less.
- 4Request the first article reportMeasured values against each marked dimension, with the instrument named. A CMM report for ±0.005 mm work, not a caliper sheet.
- 5Inspect the part on arrivalCheck the critical dimensions, the finish against the Ra callout, and the deburring on edges the drawing did not dimension.
- 6Ask about the run behind your partHow many pieces were measured, what was the spread, and was anything reworked. Specific answers signal a tracked process.
- 7Then place a small batchTen to fifty pieces exposes scheduling and repeatability, which a single part cannot show. Compare the batch report to the first article.
Questions buyers ask before a trial order
How long should a trial part take?
For a straightforward machined part, quotation and DFM feedback should arrive within a day, and a single piece can often ship in 3–5 days once the drawing is frozen.
If the part needs a special finish or heat treat, add the outside process to that estimate. Ask the supplier to name which step drives the date rather than giving one total.
Do I need to pay for a first article report?
Treat the report as part of the part, not an add-on. If a supplier charges separately for basic dimensional documentation, factor that into the comparison.
More detailed reports, such as full layout inspection on every dimension, are reasonable to charge for because they consume CMM time.
What tolerance can I expect on a large part?
Tolerance tightens as part size drops. A shop holding ±0.005 mm on a small aluminum component will not hold the same band across a 4,000 mm frame.
Mark the critical dimensions on the drawing and let the supplier quote tolerance per feature. A single blanket number across a large part is a warning sign.
Is a low quote a bad sign?
Not by itself. A low quote becomes a problem when the scope is thinner: no finish callout, no inspection, no deburring, loose packaging.
Compare quotes line by line against the same drawing revision before you decide the cheaper one is better.
Can I test a shop without sending my full design?
Yes. Use a representative test part that carries the same tolerance, material family and finish as your real part, but is not the product itself.
Sign an NDA before sharing production files. A supplier who offers one before you ask is usually set up for it.
What is the biggest cause of a failed trial?
Ambiguous drawings. Missing datums, an undefined finish, or a tolerance applied to a non-critical face all push the shop to guess.
Fix the drawing first, then judge the shop. Half of what looks like a supplier failure is a specification gap.
Send one drawing. See what comes back.
We return a quotation with free DFM analysis within 12 hours, run your trial part on the same equipment as production, and ship with a 100% inspection record.
12-hour quote100% inspectionNo MOQNDA on request
More from the shop floor
We publish setup notes, tooling trials and inspection data from the factory floor.
