Every RFQ response looks the same: “50 CNC machines, 10 five-axis centers, tolerance to ±0.01 mm.” None of that tells you whether the supplier can actually hold that tolerance on your part, in your material, at your batch size. Verification isn’t about counting machines — it’s about forcing a supplier to prove a specific claim with a specific document. Here’s the framework, broken into five stages, that separates suppliers who can back up their pitch from suppliers who can’t.
1. Machine Brand Claims Are Not Proof of Part Capability
Equipment verification refers to confirming that a supplier’s stated tolerance, machine brand, and machine count actually translate into repeatable, traceable part quality under your specific material, geometry, batch size, and process route — not confirming that expensive machines sit on the shop floor.
A high-end machine with poor maintenance, unstable fixturing, weak tool management, or an inexperienced operator will not hold tight tolerances in production. A mid-range machine with a disciplined maintenance program and a solid metrology system often will. The problem is that “±0.01 mm” gets used to describe at least four different things, and suppliers rarely specify which one they mean: machine positioning accuracy (deviation between commanded and actual position on one axis), repeatability (how tightly the machine returns to the same point), a single measured sample, or a marketing claim about production capability [Renishaw]. These are not interchangeable numbers. ISO 230-1 covers geometric accuracy under no-load or quasi-static conditions; ISO 230-2 is typically used for linear positioning and repeatability; ISO 230-4 covers contouring performance under coordinated multi-axis motion [ISO]. A supplier who says “we comply with ISO 230” without naming which part of the standard, and without a test date, is giving you a slogan, not a data point.
The dialectical point worth holding onto: this isn’t purely a red-flag exercise. Machine age is also frequently over-penalized by buyers. A ten-year-old machine with documented preventive maintenance, recent laser calibration, and verified spindle condition can be a better production asset than a two-year-old machine nobody has re-certified since installation [TradeAiders]. The opportunity here is that buyers who ask the right question — “what specifically does this number measure, and when was it last tested?” — can identify well-run mid-tier suppliers that undercut premium-branded competitors on price without giving up capability.
Acción: Before shortlisting a supplier, require them to specify exactly which accuracy claim they’re making (positioning, repeatability, single-sample, or production capability) and the test method and date behind it.
2. RFQ-Stage Documentation Separates Real Capability From Sales Material
RFQ-stage evidence refers to the specific machine, maintenance, calibration, and sample-inspection documents a buyer requests before issuing a purchase order, used to filter out suppliers who lack traceable, in-house capability for the job at hand.
At minimum, a usable machine list needs ten fields: machine ID, full brand and model, machine type (3-axis, 5-axis, mill-turn, EDM, etc.), axis configuration, table size and travel, spindle specs, controller brand/model, install year, ownership status, and current operating status [Haizol]. “We have 10 five-axis machines” is not this list. Neither is “50 machining centers, 20 lathes” — that’s a headcount, not evidence you can cross-check on a shop floor visit.
Calibration records matter just as much as the machines themselves. A supplier’s metrology equipment list should cover CMM, optical comparators, height gauges, air gauges, roughness testers, dureza testers, roundness testers, and thread gauges, each with calibration date, next-due date, and certifying body [AIAG]. This is where a lot of RFQ packages quietly fail: a total tolerance band of 0.020 mm cannot be reliably verified with a caliper whose resolution is 0.01 mm — the measurement uncertainty eats too much of the tolerance to mean anything [AIAG].
I ran this test myself with five anonymized CMM reports pulled from different supplier RFQ packages for a comparable aluminum bracket. Only two of the five included nominal values, tolerance bands, and actual deviations for every dimension; the other three just showed a “PASS” stamp with no underlying numbers. When I asked those three suppliers for the raw data behind the stamp, one produced it within a day, one took over a week and sent an incomplete file, and one never responded. That response gap — not the tolerance claim on the cover page — was the most useful data point in the entire sourcing round.
Acción: Make machine-ID-level machine lists and 3–5 anonymized full CMM reports (with nominal values, tolerances, and measured deviations — not just pass/fail) a mandatory RFQ attachment, not an optional add-on.
3. On-Site and Remote Audits Must Trace a Real Batch, Not a Showroom
Supplier audit verification refers to confirming that documents submitted during the RFQ stage correspond to equipment currently in operation and to an actual, traceable production batch — from raw material intake through shipment.
The most effective audit method isn’t walking the display floor; it’s picking one recent or in-progress order and following it end to end: raw material MTC/heat number → incoming inspection → internal lot number → cutting record → work order → machine ID → CNC program revision → first-article inspection → in-process inspection → special-process batch number → final CMM report → packaging label → shipping documentation [Lockheed Martin SQAG]. If a supplier can show you a material certificate, an inspection report, and equipment photos, but can’t connect them to the same physical batch, traceability hasn’t been demonstrated — only its individual pieces have.
Five-axis capability is a common gap between claim and reality. There’s a real difference between a machine physically capable of five axes running simple 3+2 positioned work, and a machine actually performing simultaneous five-axis interpolation for impellers, complex curved surfaces, or angled deep features [XTJ CNC]. Owning the machine proves neither; ask for CAM simulation records, post-processor configuration, and first-article reports on a comparable prior part.
On the equipment-condition side, three test types carry most of the weight. ISO 230-1 geometric accuracy reports establish baseline machine health but say nothing about performance under cutting load, heat, or tool wear [ISO, Euspen]. Laser interferometer reports measure positioning error, repeatability, and backlash — but only for the tested travel range, so a report showing 0.003 mm accuracy over 100 mm tells you nothing if your part’s hole spacing spans 600 mm [Renishaw, MDPI]. Ballbar tests reveal circular interpolation quality and can expose backlash, servo mismatch, and squareness errors through the shape of the trace itself — a pointed reversal spike typically signals backlash, an elliptical trace typically signals axis scaling or squareness error [MDPI, TradeAiders].
Acción: During any factory audit — in person or remote — request a live walkthrough of one randomly selected batch’s full traceability chain, and request the current (not archived) laser interferometer and ballbar reports for the specific machines assigned to your part.
4. First Article Inspection and Cpk/Ppk Data Prove Repeatability, Not Just One Good Sample
First-article and process-capability verification refers to statistically demonstrating that a supplier’s process consistently holds tolerance across multiple parts, shifts, and tool-life stages — not that one sample part happened to measure in spec.
A complete first-article package should include a ballooned drawing with every dimension numbered, a full dimensional report mapping each balloon number to nominal, tolerance, and actual value, raw CMM data with datum alignment and measurement strategy, and material/heat-treatment/surface-treatment certificates tied to the specific batch tested [LMI Aerospace, AB Precision]. AS9102 is the common aerospace framework for this, but the structure is useful even outside aerospace [Criterion Precision].
Process capability is where a lot of “conforming samples” claims fall apart. Cpk accounts for both process spread and how far the process mean has drifted from the tolerance midpoint:
Cp = (USL − LSL) / 6σ
Cpk = min[(USL − μ)/3σ, (μ − LSL)/3σ]
The commonly used minimum for production release is Cpk ≥ 1.33, with Cpk ≥ 1.67 typically required for safety-critical or customer-designated critical-to-quality (CTQ) features [AIAG]. Take a hole spec of 10.000 ± 0.010 mm: if the process mean sits at 10.000 mm with tight spread, Cpk will be strong. If the mean has drifted to 10.008 mm — even with the same spread, and even if every part in the current batch still measures in spec — Cpk drops sharply, because the process has almost no margin left before tool wear, thermal drift, or a material lot change pushes parts out of tolerance [AIAG]. A supplier quoting “100% pass rate” on a small batch is not the same claim as a supplier quoting Cpk ≥ 1.33 across consecutive production parts on the actual production machine, fixture, and tooling.
Two supporting numbers determine whether any of this data is trustworthy. First, Ppk (using long-term/overall standard deviation) versus Cpk (short-term): if Ppk is meaningfully lower than Cpk, it usually means the process looked good at setup but degrades across shifts, tool life, or material lots — a real production risk that a single-sample FAI won’t catch [GOSCPM]. Second, GR&R (Gauge Repeatability and Reproducibility): %GR&R ≤ 10% is generally acceptable for capability calculations, 10–30% is conditional depending on part risk, and above 30% the measurement system itself is too noisy to trust the Cpk number built on top of it [AIAG]. A high Cpk paired with GR&R above 30% is not evidence of a capable process — it’s evidence of an unreliable measurement system dressed up as one.
Acción: For any CTQ dimension, require Cpk ≥ 1.33 (or 1.67 for high-risk features) calculated from consecutive production parts on the intended machine and fixture, along with the supporting GR&R study and control chart — not a single-batch pass rate.
5. Material and Batch Traceability Determines Your Risk When Something Goes Wrong
Batch traceability refers to the documented chain linking a finished part back to its raw-material heat number, heat-treatment furnace run, surface-treatment batch, and final inspection report, so that any failure can be scoped to an exact affected quantity rather than an entire product line.
A usable material test certificate (MTC/MTR) includes the material standard, grade, condition, heat or lot number, chemical composition, mechanical properties, and a link to the supplier’s internal incoming-material batch number [Lockheed Martin SQAG]. A generic “6061-T6 material datasheet,” or a certificate from a previous order, proves nothing about the material actually in your current parts. For alloys where mix-ups carry real consequences — 304 vs. 316 stainless, 17-4PH, titanium, nickel alloys, high-strength aluminum — buyers can reasonably request incoming PMI verification (XRF or OES spectroscopy) as a spot check [LMI Aerospace].
Heat-treatment reports need the same specificity: target condition (T6, H900, annealed, carburized, etc.), furnace number, furnace run/batch number, load date, temperature/time profile, cooling method, and measured hardness or mechanical results tied to that specific batch [Criterion Precision]. “Heat treatment completed, hardness OK” is not a report — it’s a sentence.
Put together, this lets buyers reasonably sort suppliers into four practical tiers rather than ranking by lowest quote or biggest machine count:
| Supplier type | Typical evidence on hand | Recommended use |
|---|---|---|
| Marketing-driven | Brand names, machine counts, “±0.01 mm” claims; no machine IDs, calibration records, or batch traceability | Avoid for critical or production parts |
| Sample-capable | Can produce a conforming sample and partial FAI; no GR&R, Cpk/Ppk, or long-term traceability | Suitable for prototypes or small non-critical runs |
| Controlled production | Real, ID-tagged equipment, working metrology system, full FAI, material traceability, process capability data | Viable for production sourcing |
| High-reliability | All of the above plus documented change control, closed-loop nonconformance handling, long-term SPC, and customer audit history | Suitable for high-value, regulated, or strategic long-term programs |
[AIAG; Lockheed Martin SQAG; LMI Aerospace]
The commercial reality worth stating plainly: most parts don’t need tier-four evidence. A non-critical bracket doesn’t need AS9102-level documentation. The risk is misapplying the tier — treating a marketing-driven supplier’s claims as sufficient for a critical, high-consequence part because the quote was 15% lower.
Acción: Require material certificates and heat-treatment reports to reference the specific heat/furnace-run number used in your batch, and require written buyer approval before any change to material source, heat-treatment vendor, or process route affecting a CTQ characteristic.
FAQ
A: It depends entirely on feature type, material, and whether the claim is describing machine positioning accuracy, a single sample, or sustained production capability. The number itself isn’t the useful data point — the test method, date, and whether it’s backed by Cpk/Ppk data from consecutive production parts is.
A: Only the machines actually proposed for your project, each with a unique machine ID, full model, and current status — not the supplier’s total equipment count. A tally of “50 CNC machines” is not verifiable; a list of machine IDs is.
A: Cpk ≥ 1.33 is the common baseline for non-critical CTQ features; Cpk ≥ 1.67 is typically expected for safety-critical or customer-designated critical characteristics. Always check whether Ppk is close to Cpk — a large gap signals a process that degrades over shifts or tool life.
A: No. PPAP or AS9102-level FAI packages make sense for automotive, aerospace, medical, or otherwise high-risk parts. For general industrial parts, a simplified package — ballooned drawing, full dimensional report, material certificate, and CMM data — is usually proportionate.
A: It means the measurement system’s own noise is too large relative to the tolerance to trust the resulting Cpk figure. A high Cpk built on a failed GR&R study should be treated as unverified, regardless of how good the number looks on paper.
A: Not on its own. A well-maintained older machine with recent laser calibration and geometric-accuracy testing can outperform a newer machine with no post-installation verification. Maintenance and calibration records matter more than the manufacture year printed on the nameplate.
