Here's a take that surprises a lot of people in the HVAC space: most warranty failures aren't manufacturing failures. They're verification failures. Somewhere between the production line and the customer's living room, nobody checked — and a defect that should have been caught in an afternoon became a callback six months later.
I'm the quality and brand compliance manager at Empire Comfort Systems. I review every gas fireplace, thermostat, and replacement part batch before it gets the green light — roughly 200+ unique SKUs a year. In our Q1 2024 quality audit, I rejected 9% of first deliveries because they didn't meet spec. The vendors who got rejected weren't happy. That's fine. My job isn't to be liked by vendors. It's to make sure a contractor who installs an Empire Comfort Systems gas fireplace in Ohio gets the same result as one installing it in Oregon.
What Quality Inspection Actually Looks Like
The common image of a quality inspector is someone at the end of a line, eyeballing units and stamping "approved." Some of this job is still that. But gas-fired equipment requires deeper checks: ignition cycles, valve timing, thermostat calibration, gas flow consistency, and the materials and clearances that keep an installation safe.
When a batch arrives, we pull units at random and run them through documented test sequences, comparing results against our reference specs. If anything drifts outside tolerance, we investigate — and if we reject, the vendor redoes the work at their cost. In an average week I review 30–40 documentation packets and sign off on 15–20 SKUs. The paperwork is boring. Boring is the goal.
The $22,000 Thermostat Batch
In Q1 2024, we received a batch of 600 thermostats whose temperature sensors ran about 4°F high at a 68°F setpoint. Our tolerance is ±1°F. The vendor pushed back, saying it was "within industry standard."
Technically, they had a point. And I didn't care. "Industry standard" is the lowest bar anyone can clear, and it's usually the phrase you hear right before someone asks you to accept a redo on their timeline instead of yours.
We rejected the batch. They redid it at their cost — it took about three weeks, or rather closer to four when you count the revision cycle. Now every contract includes a calibration verification clause tied to our tolerance, not theirs.
What still bothers me is how close we came to missing it. The defect was invisible when a thermostat sat on a bench at a static temperature. It only showed up during a 30-minute simulated cycling test: power on, call for heat, cycle, repeat. The sensor drifted late in the third cycle. If we'd still been spot-checking the way we did in 2021, we would have signed off and shipped 600 bad thermostats.
Digital Testing Didn't Slow Us Down. It Made Us Faster.
People assume adding quality checks slows down delivery. In our case, the opposite happened.
In 2022, I implemented a digital test rig that runs every gas valve through 100 automatic ignition cycles and logs the timing curve for each one. The output is a graph so boring it could double as a sleep aid. It's also the best quality decision we've made.
Before the rig, valve testing took five days and relied on inspectors who learned to judge "bad" behavior by feel. After the rig, the same testing takes two days — and the data catches issues no human can feel, like a valve that starts sticking only after cycle 62. In its first year, the rig caught a recurring gas valve defect across two production runs that had slipped past human inspection. The defect was intermittent, so a simple functional check would have missed it. It would have failed months later, inside someone's fireplace.
So glad we built that rig when we did. We almost skipped it to hit a shipping deadline. Retrospectively, that would have cost us far more than a delay (note to self: never let a shipping deadline override testing protocol).
Replacement Parts Are Where Trust Goes to Die
New customers judge us on the fireplace. Returning customers judge us on the replacement part. And that second judgment worries me more.
When someone buys an Empire Comfort Systems thermostat to replace the one that came with their gas fireplace, they expect it to behave exactly like the original. Every time a vendor changes a sub-component, that expectation gets stressed. The most frustrating part: vendors changing internal specs without telling us, even though written specs should make that impossible. Honestly, I'm not sure why some vendors would rather argue about a spec than fix the process. My best guess is their internal QC is an afterthought.
In 2023, a vendor changed the internal spring tension on a replacement gas valve and didn't disclose it. The valve looked identical, threaded in perfectly, and opened and closed fine. But the heat-up curve was different — the fireplace took noticeably longer to reach the setpoint. To a technician, it's a minor quirk. To a homeowner, their reliable heater suddenly doesn't work right.
That issue cost us a $22,000 redo and delayed a product launch. Now we require vendors to disclose any sub-component change, and we re-verify a part whenever its bill of materials changes — even if the change looks harmless. (I should have caught that one earlier. Live and learn.)
Field Mistakes No Warranty Can Fix
Not every problem is a manufacturing problem, and I try to be straight about that with contractors and distributors.
One recurring issue: installers using foil board — foil-faced foam insulation — as a heat shield near a vent-free gas fireplace. Foil board has a temperature limit, and it's often lower than what a vent-free unit generates. We ship clearances and approved materials with every order, but we can't be there when the mantel goes up. So we test against the materials we approve and spell out the restrictions clearly in the manual.
Another confusion: fire glass. It looks like ordinary glass, and our tech support team regularly gets asked whether any glass can go into a gas fireplace. No. The tempered glass media sold for fire features is manufactured to survive high heat. Standard glass — like a beer bottle or a decorative glass bottle from a craft fair — isn't rated for that, and it can crack under thermal cycling. It looks fine on day one. Day 80 is the problem.
Per FTC advertising guidelines (ftc.gov), we also avoid claims we can't substantiate: no "maintenance-free" promises, no loose "eco-friendly" language, no efficiency numbers that don't come from our own test data. That's a compliance issue, but it's also a quality issue. If we won't verify a claim, we shouldn't make it.
The Objection: "You're Slowing Us Down"
"Quality inspection adds days to delivery."
I've heard this for four years, from operations colleagues and vendors alike. It's true. Testing takes time, and time on a production schedule is money.
What that argument misses is the cost of not testing. One field failure is a returned unit, a technician's paid hour, a customer who posts about the experience publicly, and a re-verification of the rest of our inventory. After we upgraded our verification protocol in 2022, customer satisfaction scores rose 34% and warranty returns dropped roughly 12%. I'd rather explain a rejection to a vendor than explain a shipped defect to our CEO.
I should also draw the boundary of this argument: it works for Empire because we're a mid-sized company with standardized product lines and predictable ordering patterns. If you run a small hearth shop doing one-off custom installations, the economics are genuinely different. If you're dealing with international sourcing, there are probably factors I'm not aware of. But if you sell standardized gas heating equipment and you're not verifying incoming batches, you're carrying risk you didn't price into the product.
Rejections Today Build Trust Tomorrow
When a customer buys an Empire Comfort Systems gas fireplace or thermostat, they're making a prediction. They're predicting that in three years, the pilot will light, the sensor will read accurately, and the valve will cycle exactly as it should.
My job is to make that prediction as reliable as it can be. So I'll keep rejecting the batches that don't measure up — 9% of first deliveries in Q1 2024, and more if that's what it takes. The rejection today is what makes the brand worth trusting tomorrow. That's not a flaw in our process. It's the entire point.
