Application note

I've Blown $23,000 on Measurement Equipment That Wasn't Broken. The Problem Was How I Verified It.

Posted on 2026-11-10 by Hanae Okamoto

I handle instrumentation procurement and verification for a mid-size inspection team. Eight years, hundreds of purchase orders, and a running list of mistakes I started documenting after the first expensive one.

Fourteen significant errors so far. Roughly $23,000 in wasted budget. And here's the uncomfortable part: almost none of them were caused by bad instruments.

They were caused by me. More specifically, by my assumption that an instrument that arrives calibrated, boxed, and expensive is an instrument you can trust.

Verify before you rely. Every time. It's cheaper than explaining why your data was wrong.

That sounds obvious. It isn't. I ignored it for years, and I paid for it in rework, delays, and one very awkward client call.

Argument 1: A good review doesn't tell you to check your own settings

In 2021 we bought a FLIR One Gen 3 after I read what felt like every FLIR One Gen 3 smartphone thermal camera review on the internet. The reviews were accurate about the hardware. Good sensor, decent app, solid value for the money.

What no review told me: the camera doesn't verify your emissivity settings, your distance-to-spot ratio, or whether the reference surface you're comparing against is actually at the temperature you think it is.

We used it on a batch of electrical panel inspections. Our report flagged a few "hot spots." The client's own team ran a second pass with their gear. The numbers didn't match.

Not because the FLIR was wrong. Because our settings were never checked against a known reference before we started.

Result: a $1,200 redo, a delayed maintenance window, and a client who now asks for our methodology before they ask for our price.

We had stocked up on FLIR multimeters around the same time. Same lesson, different tool. A professional-grade meter gives you a number. It doesn't give you a number you've confirmed against a second source. We added a monthly cross-check against a reference meter.

Ten minutes. Boring. Effective.

Argument 2: 'Absolute' means it knows where it is — not that it's right

The second mistake came in 2022. We installed a new CMM machine with an absolute encoder on each axis. The installers signed off. The calibration certificate went into the folder. We started measuring parts the next morning.

Three weeks later, dimensions were drifting about 0.008 mm outside tolerance on a production lot. Everyone assumed the machine was faulty.

It wasn't. The encoder zero point hadn't been verified against our datum after installation, and the room's afternoon temperature swing was outside the machine's spec window. The absolute encoder did exactly what it's supposed to do — it knew its position. It didn't know whether that position matched our reference.

Repacking and reworking that lot: $4,500. Three days of downtime. A root-cause meeting that could have been a five-minute checklist.

Here's the line I now keep taped inside the cabinet: an absolute encoder doesn't mean absolutely right.

Argument 3: 'How often should I change it' is often the wrong question

This one is counterintuitive, so stay with me.

People ask me "how often to change your columns HPLC Agilent" expecting a number. Every 500 injections. Every three months. Something fixed.

I used to run exactly that. We followed a strict replacement schedule for two years. Then in 2018 I noticed we were replacing columns that were still performing fine — while two real problems, a weakening pump seal and a slow retention-time drift, sat hidden behind "it's probably just time for a new column."

The data said replace on schedule. My gut said the schedule was masking the actual failure modes. I switched us to trend-based replacement: track peak shape, pressure, and retention time every run, and replace when the trend moves — not when the calendar says so.

I don't have hard industry-wide data on how much this saves. What I can say anecdotally is that our column spend dropped about 18% over the next year, and failed runs dropped noticeably. More important, we started catching pump issues earlier because we were actually looking at the data instead of checking a box.

Here's the uncomfortable truth: a fixed schedule feels like prevention. Sometimes it's just a delay.

You're probably thinking: 'Our equipment comes calibrated. Why re-verify?'

Fair question. I used to ask it myself.

Factory calibration confirms the instrument met spec at the factory, on the day it shipped, in the factory's environment. It does not confirm what happened in transit. It does not confirm that the install matches your use case. And in calibration terms, "traceable" means an unbroken chain of comparisons back to a national standard like NIST — it doesn't mean your specific measurement, in your specific room, on your specific sample, is correct.

You might also say your team is too small for this. I get it. My first rework cost us three days. A basic incoming-check list takes ten minutes and has caught more than 40 potential issues in the eighteen months since I built it.

Ten minutes versus three days. That's the whole argument.

What I actually changed

I formalized three things. None of them are clever.

  • A one-page incoming check for every instrument, verified against a known reference, with a signed date.
  • A monthly cross-check of handheld meters and thermal tools against a reference standard.
  • Trend logs for anything with a wear component — columns, seals, sensors — reviewed weekly.

When I first rolled this out, I second-guessed it for two months. It felt like bureaucracy. What if the checks were just theater? The doubt faded when the first real catch showed up in month three — a meter that had been drifting since delivery. Nobody would have noticed until a client report got challenged.

The bottom line

None of my expensive mistakes were about buying the wrong brand. They were about trusting a box more than a five-minute check.

If you take one thing from this: the cheapest quality control you will ever run happens before the instrument touches a job. Verify first. Then trust. Not the other way around.

I learned that the hard way — fourteen times, $23,000 later. You don't have to.

Hanae Okamoto

Hanae Okamoto

Hanae Okamoto is a process instrumentation analyst specializing in pressure gauges, pressure transmitters, flow meters, level sensors, thermocouples, temperature transmitters, and differential-pressure devices. She applies IEC 61298-2 performance methods and IEC 60751 temperature-sensor characteristics when comparing accuracy, turndown, repeatability, drift, response time, and installed uncertainty. Her work supports instrumentation engineers and plant teams choosing measurement ranges, process connections, calibration intervals, and control-loop interfaces for real operating conditions.