Your CMM Is Lying to You. So Is Your Pipette and Your Resistance Tester.

One afternoon in early 2024, a supplier called me with a problem I'd seen too many times before. Their CMM showed every part inside tolerance. Their customer showed every part out of tolerance. Both measurements couldn't be right — yet the people on both sides were convinced their data was the honest one.

So I asked the question that usually ends this kind of call: When did you last verify the machine?

There was a pause. The CMM had been moved to a different building six months earlier. Not a delicate, climate-controlled move either — it went into a bright shop floor near a loading bay door, where temperatures swung by several degrees over every shift. Nobody re-verified it after the relocation. Nobody checked whether the new environment was inside the machine's operating specifications.

That small oversight turned into 340 scrapped parts, a missed delivery deadline, and a customer who started auditing every process before trusting a single certificate again. The worst part? The CMM wasn't broken. It was just believed to be accurate. Nobody had tested that belief after the move.

For context, I'm a quality compliance manager at an industrial metrology company. Before any instrument — a workshop gauge, a laser tracker, a hexagon CMM — goes out on a demo, rental, or field inspection, my team reviews its verification record. We process roughly 200 unique instrument verifications a year. In 2024, I rejected about 11% of the first-round verification packages we received from outside labs, vendors, and internal technicians. Not because the instruments were all bad. Because the evidence of accuracy was missing, incomplete, or stale.

I've watched this pattern repeat across the whole range of measurement equipment — not just CMMs, but handheld resistance testers and laboratory pipettes. The hardware is different, the scale is different, and the cost of failure is different. The underlying problem is identical: we trust instruments because they display numbers with confidence. The instrument, meanwhile, has quietly stopped earning that trust.

The surface problem: we trust certificates more than we trust measurement

Nobody wakes up and decides to ship bad parts. What happens is more subtle. A quality engineer sees an in-date calibration certificate on a CMM, so they assume the machine is measuring correctly. An electrician grabs an insulation resistance tester with a valid calibration sticker, so they assume the reading is meaningful. A lab technician picks up a pipette that was last cleaned by the previous shift, and assumes the dispensed volume is the displayed volume.

The certificate, the sticker, and the assumption are doing the same job. They're standing in for something that hasn't actually been checked: is this instrument still measuring what it claims to measure?

That gap between paperwork and reality is where the trouble starts. I'm not exaggerating when I say I've seen precision instruments deliver confidently wrong data for months before anyone caught on. In one case, a machine shop had been using the same CMM to qualify work-in-process for a year. The annual calibration came back fine, so the team never questioned the daily results. But the calibrator hadn't checked the machine at the temperature the shop was actually running it. That single blind spot cost the company a $60,000 rejected batch.

The most dangerous error isn't the one that makes the instrument scream. It's the one that makes it whisper plausible numbers that are just slightly wrong.

Why instruments drift: the causes we conveniently ignore

Precision measurement equipment is more fragile than its industrial appearance suggests. That's not a criticism of the manufacturers — including my own. It's a physical reality that any quality professional needs to plan around.

There are four causes of drift that I see repeated in audit findings and failed verifications, and each one is predictable enough to prevent.

1. Environmental change after installation

Most CMMs are designed around a reference temperature of 20 °C. The machine itself can compensate for scale temperature, but the workpiece can't always be compensated — and the behaviour of the machine changes when one side of a room receives direct sunlight while the other side sits near a dock door.

Steel expands roughly 11.5 micrometres per metre per degree Celsius. On a one-metre part, a temperature difference of just 2 °C between the part and the machine's reference creates the same size error at the micrometer level as a whole host of mechanical problems. The CMM still works. The readings still look reasonable. They're just not the same readings you'd get in a controlled environment — and if your acceptance criteria were defined in a controlled environment, you'll accept bad parts or reject good ones without ever understanding why.

This is what I call the quiet denial phase. The instrument was right yesterday, so it must be right today. In reality, the last physical event changed the entire measurement condition — and nobody recorded that event as relevant.

2. Contamination that affects the measurement itself

Here's where the resistance tester and the pipette start telling the same story as the CMM.

A dirty pipette isn't a hygiene problem. It's a volumetric accuracy problem. If residue builds up in the tip cone of an Eppendorf pipette, it can change the surface tension behaviour of the liquid being aspirated. The result is that the pipette still clicks at the same setting, still feels normal, and delivers a slightly different volume than it says it does. At 1,000 µL the error might be negligible. At 10 µL or 2 µL, contamination can turn a perfectly calibrated pipette into an unreliable one in a matter of weeks.

Resistance testers have their own version of contamination. Test leads with cracked insulation, dirty terminals, or internal moisture from being stored in a damp workvan — all of these corrupt the measurement before the display ever lights up. A 1,000 V insulation test is only as good as the conductor leading to it. The instrument can be perfectly calibrated at the factory while the measurement path to your fingers is compromised.

3. Calibration intervals that outlive the instrument's condition

I once reviewed a resistance tester that had a valid calibration certificate from eight months earlier. The cert was fine. The instrument was not. Someone had dropped it off a ladder, and the internal reference — the component the tester uses to validate itself — had shifted enough to make every reading questionable.

Calibration certificates are a condition at a moment in time, not a promise for the future. They're only trustworthy if the instrument is used honestly, stored sensibly, and reverified after any significant event. Dropping a CMM is obvious. Moving a CMM? Not obvious. Leaving a pipette on a warm lab bench upside down? Almost invisible. Each of those events invalidates the certificate more effectively than the passing of time ever does.

4. The maintenance budget gets cut first

The last cause is the one nobody wants to admit during a procurement review. When money gets tight, verification and maintenance are seen as soft costs. They're not. They're the cheapest insurance you will ever buy against a measurement decision you cannot defend.

I've seen companies happily spend $300,000 on a CMM and then refuse a $9,000 service contract because "the machine is still under warranty." The warranty covers failures. It doesn't cover drift. The machine can sit within its warranty period and still measure badly enough to reject $40,000 worth of good parts. The warranty doesn't test the machine. Nobody does. That's the problem.

What bad measurements actually cost: the math nobody does

It's tempting to frame this as a technical issue, but it's really a financial one. The cost of an unverified measurement isn't just the cost of the calibration. It's the cost of every decision made on top of the bad reading.

Here's a calculation that converts people's understanding. Take a simple scenario that most manufacturing facilities will recognise:

• A shop runs a batch of 1,000 machined parts.
• Each part carries about $18 of material and machining cost before inspection.
• The CMM used for final inspection has drifted by just 0.014 mm on a straightness tolerance of 0.012 mm.

That's an error too small to notice and too large to ignore. If the drift goes one direction, 1,000 parts are certified as good when they're actually out of spec. The customer detects the issue at incoming inspection, returns the whole batch, and reinstates a sorting process that costs another $30 per part. The total damage is now more than $48,000 on a single CMM that nobody thought to verify after moving it across the plant.

The worst part is that the original calibration was probably fine. No one lied. No one cut a corner with malicious intent. The system simply assumed accuracy was a permanent property of an expensive machine, and that assumption turned a $2,000 verification visit into a $48,000 failure.

I've also seen this reverse direction. A lab ran parallel gravimetric checks on pipettes that were "obviously fine" because they'd been freshly autoclaved and looked spotless. The pipette with a tiny scratch inside the polypropylene tip cone failed its ISO 8655 gravimetric verification by nearly 8% at the 100 µL setting. Eight percent is not a rounding error. It's the difference between a valid experiment and a published result that cannot be reproduced by anyone else on Earth.

The same logic applies to a resistance tester used for safety verification. If the tester under-reads an insulation resistance value of 0.4 MΩ when the real value is 1.2 MΩ, you might condemn a perfectly safe cable and replace it unnecessarily. Cost: one unnecessary replacement. But if it over-reads a 0.2 MΩ insulation value as 1.5 MΩ, you'll certify a dangerous cable as safe. That's not a cost you can put on a spreadsheet. That's a liability event.

Why the cheapest quote is the most expensive strategy

Early in my career, I ran supplier evaluations the way many purchasing departments still do. I gathered three quotes, compared unit prices, and gave the job to the lowest bidder — assuming everyone was offering the same measurement capability. The three failed verifications I reviewed in 2024 all came from procurement decisions that prioritised price over evidence.

Here's the thing I had to learn the hard way: measurement equipment is not a commodity. Two CMMs with identical looking specification sheets are not the same machine in practice. Two resistance testers with the same accuracy class can behave completely differently in a dusty production environment. Two pipettes adjusted to the same volume can drift apart within months depending on how they're handled, cleaned, and stored.

My opinion now is blunt: if price is the only variable in your instrument purchasing decision, you are not buying a measurement instrument. You're buying a liability with a display screen.

That doesn't mean you should ignore price or routinely overpay for features you don't need. It means the total cost of ownership has to include things like calibration support, training, documented verification procedures, and the availability of service when the instrument drifts out of spec.

The lowest quote feels good on the day you sign it. It stops feeling good when a $300 saving on a pipette verifier leads to $14,000 of repeated experiments in a pharmaceutical lab. I've seen this equation work out badly on instruments ranging from $100 handhelds to $250,000 CMMs. The price difference between vendors was almost never the deciding factor in the total amount of money the project lost.

What actually works: verification, maintenance, and honest cleaning habits

I'm not allergic to practical recommendations. After all the diagnosis, there are concrete actions that prevent most of these failures — and they don't require a massive budget.

Verify after significant events, not just on an annual calendar

A fixed annual calibration interval is a starting point, not a complete plan. The instruments that fail most catastrophically are the ones that experience a significant event between calibrations and then continue measuring as if nothing happened.

Put a simple rule in place: any CMM, resistance tester, or pipette that is moved, dropped, exposed to extreme temperature, or involved in an incident gets a verification before the next critical measurement. That doesn't mean a full calibration every time. For a CMM, it can be a quick check against a traceable reference artefact. For a resistance tester, it means testing on a known reference before use. For a pipette, it means a gravimetric check of three or four volumes using a balance accurate enough for the task.

The trigger doesn't need to be expensive. It needs to be consistent. The documentation of that verification — clean, dated, signed, with reference values — is what will save you in an audit or a customer dispute.

Control the environment or measure its effect

If a CMM sits on a shop floor with uncontrolled temperature swings, you have two options. Either move it to a controlled environment, or characterise how the environment affects measurements during each shift. Many modern hexagon products include temperature compensation systems that help — but compensation is not magic. It reduces measured error; it doesn't eliminate the risk of an unrepresentative sample.

For electrical test equipment, the environment is about moisture and mechanical abuse. A 5 kV insulation resistance tester stored in a cold vehicle overnight can develop internal condensation that compromises high-voltage measurements until the unit dries out. Let the instrument acclimatise before you use it. That advice costs nothing.

How to clean an Eppendorf pipette (without ruining it)

Because contamination is one of the silent killers of pipette accuracy, and because the search term deserves a real answer, here is the short version of a proper cleaning routine:

• Disassemble the lower end only. Remove the tip ejector and lower shaft according to your pipette manual. Never fully disassemble the piston mechanism unless the manual explicitly says you can.
• Rinse carefully. Rinse the lower shaft and tip cone with distilled water. Do not let liquid run up into the body of the pipette unless your model is designed for full immersion — most are not.
• Wipe the exterior. Use 70% ethanol on a lint-free cloth to decontaminate the outer surface. Avoid soaking the area around the volume adjustment dial.
• Air dry upright. Leave the components to dry completely before reassembly.
• Autoclave only if the manual says so. Many Eppendorf pipette components are autoclavable at 121 °C for 20 minutes, but you must check the manual for your specific model. Autoclaving non-autoclavable parts will destroy the pipette's calibration.

That routine isn't optional laboratory politeness. It's a direct contributor to measurement integrity. A clean pipette is a verifiable pipette. A dirty one is a lottery ticket for every result you use it to produce.

Buy instruments based on verification, not just spec sheets

Finally, when you're selecting any instrument — an entry-level handheld meter or a multi-axis CMM — ask the supplier two questions before you ask about price:

1. What is the traceable verification procedure for this instrument, and can I see evidence of it?
2. If the instrument drifts in service, what does the support and reverification process look like?

This applies as much to hexagon products as it does to any other manufacturer in the market. A good company will answer those questions quickly and in writing. A mediocre one will change the subject to cost savings. The question isn't whether you can find a cheaper box of electronics with similar metrology specs. The question is whether you can prove — months later, under audit pressure — that your measurements were correct. My rule after six years of reviewing verification data is simple: if you can't prove it, you don't own it.

The bottom line

Precision instruments fail quietly. That's not a defect. It's the nature of measuring things — the closer you look, the more you realise how much of measurement is confidence, and how quickly confidence decays without evidence.

I've watched the same mistake destroy budgets across CMMs, resistance testers, and pipettes. It was never the initial purchase that caused the damage. It was the assumption that the purchase was the end of the quality process, when in fact it was only the beginning.

The lowest price looks good on the invoice. The most expensive problem is the one you trusted without proof. Verify your instruments after events, control their environment, clean them properly, and treat price as a small part of a much larger calculation. That's the belief I've come to hold after reviewing hundreds of verification records — because every interesting failure starts when someone stops asking what the instrument is actually telling them.