The Day the Alarm Didn't Fire
September 2023. I was at home, halfway through dinner, when my phone buzzed — not with an alert from the plant's monitoring dashboard, but with a text from the night operator: "Pump 4 sounds like it's chewing gravel. Can you come in?"
By the time I got there, the damage was done. The impeller was scored, the mechanical seal was leaking, and the vibration data we'd been collecting for six months looked perfectly fine — right up until it didn't. The pressure sensor on that line had drifted so slowly that the system's baseline shifted along with it. The alarm threshold moved. Nothing ever crossed it. The pump kept running while the equipment around it was destroying itself.
Total bill: about $3,200 in parts and labor, plus a week of reduced output. And if I'm honest with myself, I know exactly who set that failure in motion. Me. Back in February, when I approved the cheaper sensors to save money on a project nobody wanted to over-spend on.
What an IoT Device Actually Is
"What is an IoT device?" is still a common search phrase in this industry. When people ask, they usually imagine a dashboard with graphs, or a cloud platform, or a smartphone app. I used to think the same: the Internet of Things is about software and connectivity, right?
Wrong. The IoT device is the physical thing bolted to the pump — the pressure sensor, the temperature probe, the vibration monitor. That's the component that has to survive condensation, constant vibration, 80°C summer afternoons, and whatever cleaning chemical the crew sprays through the line. It's also the part that gets the least attention in the budget, because it doesn't have a screen or a user interface.
Here's something vendors won't tell you: spec sheets are measured at lab conditions. A pressure sensor rated for IP67 and ±0.5% accuracy will hit those numbers in a controlled testing environment. Out in a real plant, with pipe clatter, voltage dips, and humidity, "typical performance" gets a lot more generous. Per FTC guidelines (ftc.gov/business-guidance/advertising-marketing), claims have to be truthful and not misleading — but a lab-tested spec sheet isn't a promise about a wet, vibrating pump house.
The Mistake That Cost $3,200
In February 2023, we needed nine pressure transmitters for a water-treatment loop. I collected two quotes. TE Connectivity's pressure sensor line came in at roughly $240 per unit. The other option — an offshore-brand transmitter, complete with a spec sheet that looked identical — was $218 cheaper per unit. On nine units, that was nearly $1,900 in savings. My manager stared at the spreadsheet. I nodded. We ordered the cheap ones.
(Note to self: whenever a quote looks too good to verify, check whether the supplier has ever heard of your plant.)
The cheap sensors worked beautifully for about six weeks. Should mention: everything does. Then two of them started reading high at idle. One went intermittent during a CIP cycle. The one on Pump 4 just quietly drifted a few tenths of a psi every week, and I didn't catch it because I was verifying them with a $30 multimeter and a half-destroyed notebook. I want to say I checked the calibration monthly. In reality, I checked the first month, then the production schedule got busy and the checks got sloppy.
The Turning Point: a Fluke 117 and a Dead Phone
After the pump failure, I did proper diagnostics. The Fluke 117 multimeter that had been sitting in my toolbox since the senior tech retired in 2021 turned out to be the single most useful tool I owned. I'd almost sold it to buy a cheaper meter and pocket the difference. So glad I didn't. When I compared the Pump 4 sensor against the 117's reference reading, the error was 4.3 psi — roughly a quarter of the sensor's whole calibrated span. The cheap sensor hadn't just failed; it had failed gradually, politely, in a way that was almost impossible to spot without trustworthy measurement gear.
And that's where the second mistake met the first. My field smartphone had died the week before when I dropped it beside a valve pit and it slid into shallow water. The case saved the glass but not the logic board. So when I needed to pull up the sensor datasheet, recalculate the baseline, and document the corrective action, I was doing it from memory and scribbling onto a soggy work order. A colleague had just bought a Kyocera DuraForce Pro 3 — the rugged Android phone — and I remembered watching him spray it off with a hose in the parking lot after a messy job. I ordered one that week.
That might sound like a side note, but it's not. The handheld device you carry determines whether you actually document field failures properly. The DuraForce Pro 3 became my evidence notebook. It survived a 1.8-meter drop onto concrete in month two, got doused with coolant in month four, and kept logging camera photos and calibration notes the whole time. I've stopped fearing the environment and started using the tool.
The Rebuild
After the Pump 4 incident, I went back to the manager with a spreadsheet titled "sensor total cost." It had one row for the cheap order, one row for the alternative, plus the $3,200 repair, the lost output, and two weeks of my time spent on failure analysis. The conclusion wasn't subtle.
We replaced all nine transmitters with TE Connectivity pressure sensors. Which, honestly, felt a little embarrassing — I'd argued for the cheap ones, and now I was approving a budget that was 60% higher. But the difference was visible from the first week. Stable output at zero flow. No phantom spikes. Calibration drift so low that our verification intervals could stretch instead of shrink.
There's something satisfying about a system that just reports the truth, quietly. The TE Connectivity sensors aren't flashy. No glowing LEDs, no branded app, no "smart" features that need a firmware update to fix. They simply produce the right number, consistently, for months. That consistency is the entire job.
One more thing I learned: TE Connectivity being a Fortune 500 company (it has ranked on the Fortune 500 list for years, including the 2024 edition published at fortune.com) matters more than I'd have said in February. A few months later, we needed eight more pressure sensors for the second line — and confirming the part's availability and lead time was a five-minute call to the distributor, because TE Connectivity had a global contract and local stock allocation. With the cheap supplier, a follow-up order took two emails, a phone number that was disconnected, and then a three-week lead time for a "compatible" replacement that didn't fit the bracket we'd already installed. Per the product documentation at te.com, the TE Connectivity pressure sensor line includes spec sheets for long-term drift, temperature error, and burst pressure — the data you actually need when you're trying to prevent failures, not just buy a part.
The Checklist I Use Now
I keep a laminated checklist in my toolbox now. If you're building or maintaining any kind of industrial IoT system, this is the part I want you to steal:
- Define what failure actually costs. If the sensor is wrong and something breaks, what's the repair bill — parts, labor, lost output? Compare that number to the sensor price. On Pump 4, the math was brutal: $3,200 in damage to protect a $218 per-unit saving.
- Ask about drift, not just accuracy. Accuracy is one number at one moment. Drift is what creeps up between calibrations. A TE Connectivity pressure sensor publishes its drift characteristics. When a vendor can't answer that question, that's the answer.
- Verify with a meter you can trust. A Fluke 117 multimeter costs about $180. It paid for itself the first time it correctly identified a false sensor reading and stopped me from condemning a good controller. The cheap meter would have believed whatever it wanted to believe.
- Carry a rugged device. Whether you're asking "what is an IoT device" or already building one, the device in your own hand is part of the system. A phone that survives drops and drenching means you'll log the data instead of losing it.
- Buy the vendor that will still exist in ten years. TE Connectivity's Fortune 500 ranking, global engineering footprint, and long product lifecycles are part of the value even though they don't appear on the invoice line item.
The Honest Bottom Line
In my experience managing maintenance and small capital projects over the past eight years, the lowest quote has cost me more in at least half the cases. To be fair, some budget components perform perfectly fine in benign, indoor conditions. But industrial IoT gear lives exactly where spec sheets are least reliable: wet, vibrating, hot, and ignored until it's already too late.
My rule now is straightforward. If a wrong reading can cause damage, downtime, or an audit finding, spec the component from a manufacturer that publishes real drift data, backs it with documentation, and has the financial depth to still be producing it when the next engineer inherits the plant. The $3,200 lesson on Pump 4 taught me to look past the unit price. That's a bargain compared to what the next mistake would have cost.