Decisions that cost us share and were worth it

Stopping a new furnace platform to fund the cold-climate heat pump

In March 2021, my first month as chief product officer, our product plan contained two major platforms competing for the same engineering capacity: a new modulating gas furnace, budgeted at US$34 million for a 2024 launch, and a variable-speed cold-climate heat pump. Furnaces produced US$520 million of revenue in 2020, about a third of the residential business, at above-average margins. Heat pumps were 36% of our residential outdoor units and growing, but our line could not serve a northern home on its coldest nights.

I recommended stopping the furnace platform and moving its budget to the cold-climate heat pump and to dual-fuel controls that let a heat pump work alongside a furnace. We would maintain our existing furnaces, keep making them for as long as customers bought them and hold their parts for fifteen years after the last unit shipped. Corliss Abernath, our senior vice president of sales and distribution, opposed the plan in writing. Our northern distributors sold more furnaces than heat pumps, and he expected some to move furnace volume to competitors. Our chief financial officer, Mirela Szabo, asked the board to see a version with both platforms funded by delaying other work; I attached it, with my reasons for rejecting it.

The board approved my recommendation in June 2021. We launched the cold-climate heat pump in October 2023, after two winters of field trials in Minnesota and Maine, and shipped 74,000 units in 2025. Heat pumps reached 57% of our residential outdoor units that year. Furnace revenue fell to US$430 million, faster than planned, and two northern distributors moved part of their furnace business to a competitor. Residential gross margin dipped 1.1 points in 2023 before recovering.

The company now reviews every heating platform against the share of homes in each climate that are likely to choose heat pumps within the product's life. What I would do differently is the dual-fuel default. We shipped controls set to switch from heat pump to gas at 35°F, and most contractors never changed it, so many homes burned gas on days when the heat pump was cheaper to run. I should have made the switchover setting a required commissioning step, set from local energy prices.

Fitting a refrigerant leak sensor to every ducted indoor unit

Federal rules required new residential systems to use a lower-warming refrigerant from 2025. Like much of the industry, we chose a mildly flammable refrigerant, R-454B, which brought new requirements under the applicable product-safety standard, including refrigerant detection and mitigation in some installations. Our engineering reading was that a detection sensor was not required for a meaningful share of our smaller ducted systems, depending on charge and configuration.

I decided in May 2023 that every ducted indoor unit and furnace coil we sold for the new refrigerant would carry a sensor and mitigation board regardless of charge. Contractors install our equipment in attics, closets and crawl spaces that no laboratory configuration fully describes, and I did not want our safety margin to depend on a calculation made during installation. The cost was about US$46 a unit, roughly US$19 million a year at our volumes. Corliss Abernath argued that competitors' smaller systems would carry no sensor and undercut us in the price-sensitive replacement market. He was right. Sensor qualification also moved our launch from June to September 2024, while distributors filled their yards with older-refrigerant equipment.

We gave up about 2.5 points of share in two- and three-ton replacement systems during the 2025 cooling season. We shipped 610,000 systems with the new refrigerant that year. Field teams confirmed 23 genuine leaks in which the mitigation board had started the blower and shut down the compressor as designed, and we have had no reports of ignition. Two production homebuilders chose us partly because every system carried a sensor.

The cost I did not anticipate was nuisance trips. Our first sensor firmware reacted to some household aerosols and solvents, and in early 2025 about 0.6% of installations recorded a false alarm, which meant service calls and anxious homeowners. A firmware change in May 2025 brought that to 0.08%. I would run a far larger in-home pilot next time: we tested in 40 homes, and 300 would have found the problem before launch.

Recalling every heat kit from a nine-month production run

In March 2023 our product-safety team brought me a second report of a fire that began in an electric heat kit, the resistance heater installed in an air handler for backup heat. There had been nine field reports in all, two fires with property damage and no injuries. The cause was a high-limit switch that could fail to open, from a supplier change in February 2022. Heat kits built between February and November 2022 numbered 52,000. About 14,000 could be traced to the supplier lot we suspected, but for seven months our Joplin plant had not recorded component lot codes at kit assembly, so we could not show that the others were safe.

Gemma Truscott, who leads product safety, and our general counsel, Trent Okafor, agreed that we had to report to the Consumer Product Safety Commission. The disagreement was about scope. Our chief financial officer and our head of distribution favoured a recall of the 14,000 traceable kits, with a service bulletin for the rest. I recommended recalling all 52,000, because a recall limited to the units we could trace would protect our records rather than the households with untraceable kits. The full recall cost US$16.8 million against an estimated US$5 million for the narrower one. Rosamund Quarrie and the board's product safety and quality committee approved it.

We announced the recall with the Commission in June 2023. Fewer than half of the kits had registered owners, so we used distributors' sales records to reach the contractors who had installed them, paid any licensed contractor US$185 for each inspection and replacement and sent letters to the addresses we had. Eighteen months later 78% of the kits had been inspected or replaced. There were no further fires among remedied units.

Lot codes for every safety-critical component are now recorded at assembly, and our connected systems report heat-kit temperatures to a monitored dashboard. What I would do differently is the nineteen days we took after the second fire report to establish the failure mode before filing. I would now file on the first credible pattern and finish the engineering alongside the Commission.

Opening our service tools to every licensed technician

When we launched our communicating control in 2018, I restricted its service app to Ambersley certified dealers, then about 2,600 of the 11,000 firms installing our equipment. Certified dealers paid for training and wanted something in return. By 2022 the consequences were plain. Independent technicians could not read fault codes on our communicating systems, homeowners whose dealer had retired could not get service, and 29% of control boards returned under warranty had nothing wrong with them. Homeowner complaints about being unable to get service had tripled in four years.

In January 2023 I proposed reversing my own decision. Service manuals, wiring diagrams and fault-code tables would be published without a login. The app's diagnostic and commissioning functions would be open to any technician holding EPA Section 608 certification, the federal credential already required to handle refrigerant, and repair parts would be sold through every distributor to any contractor. A short, published list of functions, such as factory calibration of the compressor drive, would stay restricted for safety reasons.

Our dealer advisory council, chaired by Lyle Bellweather, opposed it, and Corliss Abernath expected the certified programme to lose its value. I made a mistake in process: I brought the council a decision rather than a proposal, and they told me so. We gave up about US$3.1 million a year of programme fees tied to tool access, and certified dealers lost the automatic fault alerts that had sent them service calls.

By mid-2026, 7,900 contracting firms used the app. First-visit repairs on communicating systems rose from 71% to 84%, and parts returned with no fault found fell to 18%. Complaints about unavailable service fell by 62%. Certified dealer numbers dropped to 2,350 in 2023, then recovered to 2,540 as the programme was rebuilt around training, extended warranty registration and co-operative marketing.

Apart from the process, I would change one design decision. Contractor access to a home's data expired ninety days after each job, which irritated homeowners with maintenance agreements. We have since let homeowners name an ongoing contractor, which we should have offered from the start.