Supplier's Certificate Said Pass. Our FARO Arm CMM Didn't Agree.

The supplier's report said +0.002 mm. Our FARO Arm CMM measured -0.009 mm on the same bore. Same part. Same drawing. Two calibrated CMMs. One production line that couldn't wait while we figured out who was right.

I'm the quality manager at a precision manufacturing company. I review every first article before it reaches production - roughly 200 unique part numbers a year. In the first half of 2024, I rejected 9% of initial supplier submissions because the measurement evidence didn't survive our own verification. Not because every part was bad. Because the certificate and the reality were telling different stories.

The argument was never about the part

To be fair, most supplier quality engineers aren't trying to deceive you. The uncomfortable truth is that a certificate is a story - a measurement told under one set of conditions. Your incoming inspection is a second story told under another set. When the two disagree, we automatically assume someone is wrong. Usually, they're both right about the story they told. The missing pages are the ones nobody wrote down.

Which datum was aligned first? Which probe tip was used? Least squares or maximum inscribed? At what temperature? I've lost count of the arguments that turned out to be about these questions instead of about the actual part.

Calibrated doesn't mean indistinguishable

Here's what I learned the expensive way: a calibration certificate says the instrument performed within spec on a particular day, under defined conditions. It doesn't say that every measurement made after that day inherits that spec.

Standards like ISO 10360 for CMMs and ASME B89.4.22 for articulated-arm CMMs define accuracy under controlled conditions. In a 20 degrees C lab, with clean fixturing and a skilled operator, a machine delivers numbers near its spec sheet. On a shop floor at 28 degrees C, with different tooling and a different operator, the same machine can deliver a perfectly valid measurement that doesn't match the first one. Neither measurement is wrong. They're just not identical.

I assumed same specifications meant identical results across measurement systems. Didn't verify. Turned out each system - even two machines from the same manufacturer - had slightly different habits once you changed the setup. Learning not to assume that saved us far more than any calibration contract.

The same breakdown happens outside dimensional metrology

Once you see this pattern, you'll see it everywhere. A calibrated tester can still give you garbage if the things around it are bad.

Example. A 1507 insulation resistance tester can display a confident number, but if the multimeter test leads attached to it have an intermittent internal break, the meter might report high resistance when the circuit is actually open - or fluctuate when you wiggle the lead. The instrument is calibrated. The measurement chain is not.

Same with thermal imaging. The camera sensor can be flawless, and the image can still mislead you if the emissivity setting is wrong. So when a colleague asked me the Topdon vs FLIR thermal camera question, I said: brand isn't the first question. The first question is how you'll verify what the camera tells you - and whether the person using it knows what the settings actually do.

What that disagreement really costs

Two years ago, we received an 8,000-unit order where the supplier's CMM report said the bore was in tolerance. Our FARO Arm CMM said it wasn't. We did what most people do: we argued.

We argued long enough that we had to expedite replacement material. The total rework, expediting, and downtime cost us about $22,000. The supplier wasn't trying to cheat us. We weren't trying to cheat them. We simply had no agreed-upon definition of what in tolerance meant, because our measurement methods were different.

What hurt most wasn't arguing with the supplier. It was realizing that if the numbers had gone the other way - if I'd accepted the certificate to keep the peace - we might have shipped an out-of-tolerance part to our customer. That version of the story costs even more, and you don't see the bill until months later. Usually from your customer, not your supplier.

We didn't have a formal process for reconciling these disagreements. That absence was the defect. The third time it happened, I finally created the verification protocol I should have built after the first. (Note to self: build the process after the first time, not the third.)

What I do differently now

1. Define the measurement before you define the tolerance. Every contract and PPAP now says how to measure: which datums, which fitting method, which probe size, how many points, and at what temperature. If a bore calls out a diameter, we also specify whether it's least squares or min/max inscribed. If I'd done this earlier, both sides could have proven their point - and we would have seen that the real issue was the specification, not the parts.
2. Ask what's NOT included. I've learned to ask that before what's the price - for measurement systems and for any quote that promises quality. A certificate is a deliverable too. Does the price include calibration, training, software, and ongoing support? A quote that lists everything upfront may look higher, but in my experience it costs less at the end. The vendor who hides fees until later is telling you something about how they handle other kinds of truth. This applies to small purchases like multimeter test leads as well: if the cheap ones aren't verified and reliable, they're not a bargain.
3. Verify the verifier first. Before we use any instrument for an acceptance decision, we check it against something we trust: a calibrated artifact, a reference resistor, a known standard. For dimensional work, our FARO Arm CMM gets a quick verification against a calibrated length standard before critical runs. It takes a few minutes. It has prevented more than one embarrassing correction.
4. Make the data visible, not just the conclusion. I don't accept a single number from a supplier anymore. I ask for the raw data and a description of the measurement conditions. On our side, we can do the same. Because FARO arm data is digital, we can bring a point cloud up on the FARO app in a meeting and ask: which points did you use to fit this plane? That question ends more arguments than any calibration certificate ever has.

If we agree on the measurement chain, most disagreements vanish. When a number does surprise us, we investigate instead of assigning blame. That's the real shift. The certificate is a claim; verification makes it confidence. Show me your measurement conditions, show me your raw data, and I'll show you mine.