Tuesday, 4:40 PM. A customer texts you a photo of their new ecobee thermostat screen with a low-battery warning. It's the third one this month in your service area. Then a homeowner calls because their refrigerator compressor is "very hot to the touch." And the job before that? An Atwood propane carbon monoxide detector that won't stop chirping, plus a dishwasher that "won't clean anything."
Before you book all four visits, ask the same question our failure-analysis team starts with: How many of these are actual equipment failures?
I'm a quality and compliance manager at ecobee. I review every returned unit and warranty claim that comes through our facility — roughly 600 unique units a year. In our Q2 2024 audit, we went through 340 returned units tagged "defective." We tested every one against spec. No fewer than 38% were fully functional. No defects. No intermittent glitches. They worked exactly as designed.
That number changed how I think about service callbacks. Most of what gets called "broken" isn't the hardware. It's a gap between how a device actually behaves and what the owner expects it to do. And for contractors, that gap is expensive — whether you bill the visit or eat the hour.
The Real Problem: We Hand Over Hardware Without Handing Over "Normal"
Here's something this job taught me that I didn't expect: you can't say whether something passed inspection unless you know what "normal" looks like. Sounds obvious. But when you review complaints long enough, you realize nobody ever hands that information to the end user.
Manuals describe features. They rarely describe what the device is supposed to feel like.
It took me about three years and four hundred returned units to understand that the most common failure is a failure of expectations, not parts. Once I started tracking it, the same patterns showed up in our data over and over. Here are three of them.
The ecobee "Battery" That Isn't a Battery
"Battery's dead, and I can't find where to replace it." That's one of the most common notes on warranty returns from people looking up "ecobee thermostat battery location." They've owned older battery-powered thermostats. They expect a door, a compartment, two AAs.
There isn't one. The ecobee has an internal backup battery that keeps settings and scheduling alive during power blips, and it's charged by the HVAC system through the thermostat wiring. You can watch the voltage in the settings menu. When a customer shows you a low battery alert, the battery isn't the story — the story is power delivery. It's usually a missing C-wire or a two-wire heat-only system where the Power Extender Kit (PEK) wasn't used.
We saw this in a return back in March 2024: a unit tagged "defective battery." Our bench team ran it for three days straight — no issues at all. The customer's system was a two-wire heat-only setup with no PEK, so the thermostat had been running entirely off its internal cell until it gave up. The hardware was fine. The installation was underpowered.
So when you see "low battery" in the ecobee app, the question isn't "where's the battery door?" It's "what's feeding this unit?" That reframing has saved contractors I work with an embarrassing callback or two.
The Atwood Detector That Won't Shut Up
Atwood propane carbon monoxide detectors — the ones in RVs and travel trailers — are a close second for misdiagnosis. Customer story: "It chirps every thirty seconds, so it's broken." Sometimes that's true. Often it isn't.
These units do two jobs: sense propane and sense carbon monoxide. They also produce two kinds of warning: an alarm for a real reading and a chirp for a system condition. People mix them up. And the condition people mix up most is end of life.
Most CO alarms, including Atwood units, are rated for five to seven years from the manufacture date. When the sensor reaches end of life, the device chirps on purpose. That's not a malfunction — it's one of the few cases where the product is doing exactly what UL 2034 expects, and the user interprets it as a defect.
Before you replace that chirping detector, look at the date code on the side. Five years old? The chirp is the unit working correctly, and swapping it out is the right move. Two years old and chirping? Different conversation — but at least you're having the right one.
The Dishwasher That "Won't Clean"
This one gets under my skin, because I've seen service reports that ended with "replace circulation pump" when the actual problem was geometry. Dishwashers clean by spraying water. If the spray arm is blocked, or bowls are nested so tightly the water can't escape, the machine "fails." Customer calls it broken. Contractor drives out, finds nothing, charges a trip fee, and both sides are annoyed.
How to put dishes in a dishwasher correctly isn't a secret, but nobody says it out loud at installation: plates face the center, where the water spray is strongest. Bowls face down, not up — they fill with water otherwise. Big items in the lower rack block the spray arm faster than people expect, and tiny items fall through and jam the impeller. The machine isn't broken. The water is being asked to reach places physics won't allow.
That's not a component failure. It's an information failure — and the homeowner pays for it.
What Misdiagnosis Actually Costs
Let's make this concrete. I'm not going to pretend I know your exact numbers, but you already know them in your gut: truck hours, mileage, the call that could've been a $1,200 install turning into a free reassessment.
Two avoidable callbacks a month works out to maybe two to four hours of lost time. At typical service rates, that's somewhere between $2,000 and $7,000 a year of earning potential — gone. And that's assuming you'll charge for the callback. A lot of contractors won't, because the customer's already annoyed.
There's a worse cost, though. When customers learn to ignore alarms because the last one was a false alarm, you've trained them to dismiss the warning that actually matters. The chirping CO detector that "does this thing all the time" might be sitting in a camper with a slow propane leak. The compressor that's "just a little hot" might be running on a dirty condenser coil, cooking itself toward a $900 replacement.
The expensive mistake works in both directions. Treating normal as defective wastes money. Treating defective as normal costs more, eventually.
I learned never to assume the complaint describes the defect after a unit came back to us tagged "no power." Our bench tech found a tripped breaker in the customer's panel, reset it, and the unit ran fine. The hardware didn't fail. The interpretation did.
The Fix: Prevention Over Cure
Define normal before you diagnose abnormal. That's the whole job — whether the device is a thermostat, a detector, or a dishwasher.
You don't need a bigger warranty budget or a faster truck to fix most of this. You need a checklist, and a little willingness to hand customers better information. Five minutes in front of the equipment prevents five days of back-and-forth:
- Thermostats: Walk the customer through the battery voltage readout in the ecobee menu (Menu → Settings → Battery, roughly — check the current ecobee smart thermostat operating instructions for your model). Say the same sentence every time: "This doesn't take disposable batteries. A low-battery alert is a power-supply question, not a battery question."
- CO/propane detectors: Write the date-of-install on the side of the unit with a sharpie. Tell the customer the chirp pattern for end-of-life vs. alarm, and set the expectation that the sensor expires in five to seven years. That's not a defect. That's the device working as intended.
- Refrigerators: Put your hand on the compressor housing while the customer watches. Say: "This is going to feel hot — anywhere from warm to about 150°F — and that's within spec. If it feels burning hot even with the condenser coils clean and clear, call us." One sentence. One fewer callback.
- Dishwashers: Show them the loading pattern while you're on site: plates toward the center, bowls face down, spray arm clear. Leave a one-page guide if you have one. Sounds kinda patronizing, I know — until you've been called back for a "broken" dishwasher that had a cookie sheet lying across the lower spray arm.
If you want to go one step further, make a "This Is Normal" sheet and leave it on every job. It takes twenty minutes to write and will prevent more callbacks than any discount you could offer.
I can only speak to my own numbers: when we redesigned our field-troubleshooting checklist back in 2022, support contacts about "device not working" dropped about a quarter. We didn't fix the hardware. We fixed the information gap around it.
That's the thing I wish I'd understood years ago. The devices mostly work. The handoffs are where they break.