A CNC supplier can be released for production only after the certificate scope is checked against your industry and material, a first article is produced on the live production setup, and an FAI package — not a generic CMM line — is reviewed.

Sample approval alone is not a sufficient gate: a trial order and a feature-level inspection plan should come first, and the sources do not settle sample counts, AQL levels, capacity-verification methods or payment terms, so those must be asked for explicitly.

Check the certificate scope before you order any sample

A certificate logo does not tell you what the quality system actually covers. The scope statement is the part that maps to your industry and material, and it is the item to verify before a sample is ordered17.

ISO 9001 covers most industrial work, while AS9100D is the aerospace standard and incorporates all ISO 9001:2015 requirements while adding aerospace-specific ones, including risk management across the supply chain15,17.

Measurement competence sits outside the QMS certificate: A2LA accreditation validates it separately, so a shop can hold a valid QMS certificate and still not have accredited measurement capability17. Suppliers publish their own certification positions, and those are claims to check rather than evidence.

Yonglihao Machinery states it holds ISO 9001, AS9100D and IATF 16949 covering aerospace, automotive and general commercial programs16. Fictiv states it supports ISO 9001:2015 and works with partners certified to AS9100 Rev D, ISO 13485:2016 and IATF 16949:20167.

Both are supplier positions; the buyer's check is the scope document behind the logo, not the logo itself17.

The first article must come off the live production setup

The first article verifies the production process, not the part. It must be produced on the exact production setup — machine, fixtures, workholding, tooling and program revision — and must not be a hand-finished prototype or pre-production sample2.

A sample made any other way tells you nothing about the batch that will follow, because the setup is what the batch will run on. FAI is the initial verification of a new or changed manufacturing process, and it is a reporting and approval scope rather than a specific measuring instrument8,9.

That distinction matters at the gate: the buyer should require evidence that the sample came from the live setup and that the FAI report covers the agreed characteristics2,8.

Request FAI when the part is new, the drawing revision changed, the material or surface treatment changed, a new fixture is used, or a critical assembly feature must be verified before a batch proceeds9. The sources do not state how many samples constitute an FAI, who must approve it, or what turnaround applies8,9.

Specify the FAI report type and the characteristics it covers

A quote line that says "CMM inspection" is not a first article. CMM describes a measurement method, while FAI and full dimensional inspection describe the scope of the report8.

A CMM report does not mean every drawing characteristic was inspected, and a basic dimension report may omit critical GD&T or datum relationships8. A full dimensional report covers all agreed drawing characteristics and may combine CMM, hand tools, gauges and roughness testing8.

When a first article is needed, ask for an FAI package or a defined inspection report rather than only "CMM inspection"9. Review the report for completeness and AS9102 compliance; missing dimensions, no ballooned drawing, or no customer approval are failure signals18.

Read the measured values against nominal on the ballooned drawing, because measurements clustering near tolerance limits signal a process running at its capability boundary before volume begins19.

Match the inspection method to each feature and set 100% versus sampling per feature

Inspection tool accuracy ranges differ by orders of magnitude, so the method has to match the feature tolerance. A caliper cannot resolve what a micrometer or CMM can, and the choice of instrument is part of the inspection plan rather than an afterthought20.

| Inspection tool | Stated accuracy | |---|---| | Digital caliper | ±0.02–0.03 mm | | Outside micrometer | ±0.002–0.005 mm | | CMM | ±0.001–0.002 mm | | Optical comparator | ±0.005–0.025 mm | Do not specify 100% inspection by default.

It can be appropriate for safety-critical, high-value or especially sensitive features, but it changes inspection time and cost9. State which features require 100% measurement, which can be sampled, and whether a first, last or random sample is required9.

Inspection intensity should be defined per feature and per order stage: a prototype may need more engineering review per part, while a repeat batch may use defined sampling if the drawing and purchase order allow it28. Neither source provides an AQL level or sampling frequency9,28.

Require an unbroken heat/lot traceability chain and a mill cert

Material grade belongs on the drawing, and the test report belongs on the material submittal. Name the grade on the drawing, name the corrosion class on the spec, and name the test report on the material submittal33.

Lot traceability should be required as a pair: a mill cert for the material and a dimensional inspection report for the part24. Traceability has to survive the whole route.

Heat and lot numbers must carry from receiving through final inspection, and in-process and final inspection records must tie to the work order and serial or lot number18. Lot mixing, missing receiving records or a broken traceability chain are the failure signals that should stop an award18.

YISHANG states it offers documentation and technical support including heat-number traceability and FAI/sample approval23; that is the supplier's own position, and the buyer's check is the documentation package itself. Neither source states the record retention period or the mill-cert format18,24.

Treat sample approval as one gate, then run a trial order

Sample approval should not be the only gate. Pilot runs, documentation review and repeat-order performance are better indicators of long-term fit11.

A trial order before committing production volume validates actual lead times, communication quality and part conformance before a full program is exposed to an unproven supplier17. The sequence is sample approval, then pilot run and documentation review, then the production release decision11.

Neither source specifies the number of samples, the approval criteria, or the size of the trial order11,17.

Confirm the lead time and the machines behind the quote

A quoted lead time is a calendar promise unless it names the work inside it. Confirm whether the quoted lead time includes first article review, correction loops and final report approval; if it does not, the production schedule promise is unprotected12.

For repeat orders, ask how tool changes, lot changes, sampling frequency and measurement records will be handled, so a prototype schedule is not reused as a production promise12. Capacity is a separate question from capability.

A shop's brochure may list a machine that is committed to other programs, so ask which specific machines will actually run your job17. A quote that arrives with no questions about tolerances or material is a red flag that the drawing was not actually reviewed17.

The sources do not provide a method to verify shifts, current utilization or quoted lead time at a target monthly volume12,17.

Normalize one-time costs before comparing total delivered cost

Setup, programming, tooling preparation, fixture design and inspection planning are spread across the order, so a single prototype absorbs nearly all of the non-recurring work into one part28.

That is why one-time costs — setup, programming, tooling, fixtures, qualification and first article — should be normalized across suppliers on the same quantity tier and unit of measure before total delivered cost is compared13.

Prototype and production requests should not be combined into one submission: prototyping prioritizes speed and DFM feedback at 1–50 units with CTQ-dimension inspection, while production quotes focus on cost stability and process repeatability with FAI, CMM, MTR and SPC data10.

A no-MOQ supplier gives ordering flexibility, but the buyer should expect a higher per-unit price because bulk savings are lost30. No source gives payment terms or specific price-break percentages at named volume tiers13,30.

Where the sources disagree

Disputed item (with unit)One source reportsAnother reportsWhat the buyer should do
Prototype per-unit cost premium (multiple of production unit price)1–10 parts at 3–5x production per-unit costNo CNC-specific multiplier publishedAsk the supplier for its own prototype and production-lot prices at your quantities
Per-unit cost curve (units where cost flattens)Flattens at 50–100 partsNo CNC volume tier givenRequest price breaks at your realistic annual volume, not a guessed tier
316 vs 304 bar stock premium (% at same diameter and finish)30–50% premiumNo comparable figure publishedConfirm the grade on the drawing before comparing quotes across suppliers

What the sources do not establish

  • No source provides a method to verify a supplier's certificate scope against a specific material or industry for a named supplier.
  • No source states the required number of samples, approval criteria, or sample turnaround time for sample approval.
  • No source provides an AS9102 report template or states who must approve the FAI.
  • No source states achievable tolerance or surface finish values for a specific material and feature size, nor measurement uncertainty values.
  • No source provides a method to verify shifts, current utilization, or quoted lead time at a target monthly volume.
  • No source gives an AQL sampling frequency or states whether the buyer or a third party may witness final inspection.
  • No source gives payment terms or specific price-break percentages at named volume tiers.
  • No source provides a mill-cert verification procedure or record retention period.
  • No source addresses the specific approval workflow (who signs off, in what sequence) between first order and batch production.
Sources · 21

Technical references cited for verifiability — not supplier recommendations. Browse the research library.