A $1,410 Test-Equipment “Savings” Cost Us $41,300 — and a Failed Audit
On August 16, 2022, I approved a purchase that I would spend most of 2023 explaining to auditors. It didn’t look risky at the time. Our RF test bench needed another power sensor, and the quote for a genuine Keysight power sensor was about $2,200 (based on the distributor quote we received in August 2022; prices have moved since). A third-party “compatible” sensor was $790. I did the arithmetic: $1,410 in apparent savings. I hit approve.
Before you judge me too harshly, you should know who does the approving. I’m the calibration and metrology lead for a contract electronics manufacturer that builds controllers for medical and laboratory equipment. I’ve been handling test-equipment purchases and calibration orders for nine years, and I’ve personally made — and documented — six significant buying mistakes, totaling roughly $174,000 in wasted budget. I keep a pre-purchase checklist so the next person doesn’t have to repeat them. This mistake is item one.
The Warning Sign I Already Had
I should have learned the lesson a year earlier, with multimeters. In 2021, our senior engineer specified a Keysight Technologies bench digital multimeter for an automated test station. I bought a lower-priced alternative instead, saved about $700, and spent the next month fighting its half-compatible SCPI command set. Then I spent another few weeks explaining why its calibration certificate didn’t include the uncertainty statement our quality system required. We replaced it with the 34461A unit we should have ordered at the beginning. The budget meter went to the repair bench, where it still works fine for “is this power supply dead?” checks. That should have told me something.
The Sensor That Wasn’t in the Database
The RF sensor problem took longer to show up. For the first month it seemed fine. We compared it against a calibrated power meter at 1 GHz and the numbers agreed within 0.1 dB. What we did not fully appreciate is that a genuine Keysight power sensor stores its calibration factors internally and the power meter reads them automatically. The third-party sensor did not. The calibration table had to be entered by hand. Someone typed it in; a sticky note appeared on the bench; the sticky note disappeared. At 5.8 GHz the reading ran about 1.2 dB high. A module transmitting below specification looked as if it were inside the limit. We shipped 214 units, and the customer’s incoming inspection found 41 with output power below spec.
When a customer rejects 41 units, you don’t just send replacements. You get a corrective action request, and that request comes with an audit. During audit prep, our quality engineer asked for the calibration record of the sensor on that test bench. I searched the asset database, then the second database, then the paper files in the back cabinet. The sensor wasn’t in any of them. It had never been sent for calibration, partly because the vendor didn’t offer an accredited calibration service for it.
On the day of the audit, the customer’s quality engineer asked three questions: Which sensor measured those modules? What was its serial number? Where is its certificate? I could answer two.
“If you can’t prove it was calibrated, you didn’t measure it.”
That sentence is the worst thing to hear when a product you shipped depends on the measurement, and I still repeat it at every training session. The one part of the audit that went well involved temperature. Our temperature work uses an old 714B thermocouple calibrator that has been on the bench for almost twenty years. It has no touchscreen and no Bluetooth. The plastic case is yellowed. But it has a current, accredited calibration certificate with a stated uncertainty, and our technician can show exactly how it is used. The auditor did not ask us to replace it. She asked why we did not manage every instrument that way.
The same audit prep included a close call with the dimensional lab. The laser tracker we use to verify our robotic assembly cell was purchased used in 2021, and its calibration certificate existed only in the email account of a technician who had left the company. It took us an hour to find. That was the moment I added “documentation before dock” to our purchasing process.
What the $1,410 “Savings” Actually Cost
All of that is how a $1,410 savings became a $41,300 failure. I’m not counting only returned goods. I mean rework, quarantine, re-validation, audit hours, premium freight, and a six-week schedule slip on one customer program. That number still appears on a slide in our quality review deck. It is a useful slide.
We ordered the replacements in January 2024: a genuine Keysight power sensor with the calibration table stored in the sensor itself, plus a set of Keysight Technologies bench digital multimeters to replace the remaining budget meters on certified stations. I’m not going to pretend that premium gear never fails—it does. But premium gear comes with an ecosystem: calibration services, drivers that behave, application support, and paperwork that survives an audit. When a measurement can end up in a report to a customer, that ecosystem is not a luxury.
The Cost-of-Ownership Checklist We Use Now
We still buy budget tools. I have a $20 multimeter in my car and I like it. But it will never measure a product that leaves the building. For anything that does, we run this checklist:
- Decide where the measurement will land. If it will appear on a customer report or a certificate, the instrument needs an accredited calibration and a documented chain of traceability.
- Check the service ecosystem before approving the purchase. Is accredited calibration available? Are drivers and documentation current? How long does calibration take, and what does it cost? If the vendor doesn’t support the instrument, the lower price isn’t lower.
- Put the calibration certificate into the QMS before the instrument is used. No record, no measurement.
- Train the person. Even a good instrument is only as good as the hands using it.
That last point is not a joke. A few weeks after the audit, I was training a new technician on the wet-chemistry bench. I asked, “Do you know how to use an Eppendorf repeater pipette?” He said yes, because he had used single-channel pipettes in university. He didn’t realize an Eppendorf repeater pipette needs to be primed before the first dispense. He was maybe 30 seconds away from preparing calibration standards with an unprimed tip. If I hadn’t asked, we would have discovered the problem the next day, after wasting a day and a box of expensive tips. Same false economy, smaller numbers.
So I haven’t sworn off cheaper equipment. I’ve sworn off assuming that the price tag is the total cost. The total cost includes calibration, support, training, documentation, and the price of being wrong in a way you can’t defend.
The cheapest instrument is the one you can prove, maintain, and defend. Sometimes that instrument is an old 714B thermocouple calibrator. Sometimes it’s a used laser tracker. And sometimes it’s the Keysight Technologies bench digital multimeter I should have bought the first time — the one whose only crime is that it costs what it costs.