On August 3, Lennox told investors residential HVAC revenue fell 7 percent in the second quarter, with unit volume down 12 percent, and pushed the meaningful recovery out to 2027. Four days earlier, Carrier reported the opposite: residential sales up 9 percent, field inventory running 25 percent below last year, and CEO Dave Gitlin saying the market is “fundamentally a replacement business” again. Both companies are reading the same market. The difference between their numbers is sitting in your service history right now, as repair tickets you closed and never turned into a replacement quote.
Two manufacturers, one market, opposite readings
Gitlin gave the line that matters to a contractor: “There’s only so long a customer can repair over replace.” Lennox CEO Alok Maskara said the same thing from the other direction. Lennox believes many of the repairs it is seeing represent deferred replacements rather than permanently lost equipment demand, and that the underlying demand profile has not changed.
Strip out the investor framing and both statements say one thing. The replacement demand exists. It is just parked inside repair calls. A manufacturer can wait for that demand to show up in 2027 because a manufacturer sells through thousands of contractors. You cannot wait, because the homeowner whose compressor you replaced in June will call whoever answers when the system dies in January. That is a coin flip on a $9,000 to $14,000 ticket you already earned the right to win.
Two more numbers make the timing specific. Lennox added Section 232 tariff pricing on July 1 and expects roughly 5 percent inflation this year. Carrier realized about 3 percent residential pricing in Q2 and expects roughly 4 percent for the full year. Equipment is not getting cheaper between now and 2027.
The list is already in your service history
This is not the dead-estimate problem. Dead estimates are jobs you quoted and lost. This is the layer underneath: jobs you completed, got paid for, and closed without ever putting a replacement number in front of the customer. The homeowner already paid you. They already trust your tech. Their system already failed in front of them. That is the warmest list in the business and almost nobody works it.
Six signals separate a deferred replacement from a normal repair. Score every completed job from the last 36 months against them:
- System age 12 years or older at time of service
- Two or more repair visits on the same unit inside 24 months
- A single repair ticket over $600
- A major component replaced: compressor, heat exchanger, evaporator coil, control board
- R-22 or early R-410A equipment, where refrigerant cost turns the next repair into a bad conversation
- No active maintenance agreement on the account
Three or more signals on one account means that customer is buying a system in the next 18 months. The only open question is who sells it to them.
Build the list this week
Export completed jobs from the last 36 months out of your CRM. ServiceTitan, Housecall Pro, and Jobber all export job history to CSV. Pull these columns and nothing else: customer name, address, phone, job date, job description, invoice total, equipment model, equipment install year, membership status. Strip anything you would not want in a spreadsheet on a laptop.
Then run the file through Claude, ChatGPT, or Gemini with a prompt that scores rather than summarizes. Something close to this:
“You are scoring a service history file for deferred replacement opportunities. For each unique service address, count how many of these apply: system 12+ years old, 2+ repair visits within any 24-month window, any single invoice over $600, replacement of a compressor / heat exchanger / evaporator coil / control board, R-22 or pre-2018 R-410A equipment, no active membership. Return a table sorted by score descending with columns: address, customer name, phone, score, signals hit, most recent job date, total repair spend across all visits. Flag any account whose total repair spend across the file exceeds 40 percent of a typical replacement cost of $9,000. Do not invent data. If a field is missing, mark it UNKNOWN.”
The total-repair-spend flag is the one that does the selling. When a homeowner sees they have already put $3,800 into a 15-year-old system, the replacement conversation stops being a pitch and becomes arithmetic.
Cost of doing this: a $20 to $30 monthly AI subscription you probably already pay for, and about two hours of somebody’s afternoon. That is the entire stack. Voice AI vendors will sell you the outbound calling layer on top, and that market is real money now. Avoca raised $125 million at a $1 billion valuation in April 2026 and went from 10 customers in 2024 to more than 800. But do not start there. Run the list manually for 30 days first, so you know your own conversion rate before you automate a process you have never actually performed.
Price the urgency honestly
ACHR News published the August 2026 price increase list on August 5, and it is broad. Amana, Goodman, and Daikin equipment went up as much as 7 percent effective August 3. Legend Valve ran 3 to 25 percent depending on category on August 1. RectorSeal added 3 percent on August 6. Emerson Nidec Motors raises on August 16. Lau parts jump 16.2 to 16.8 percent on August 31.
Here is where most shops overplay the hand. On a $9,000 install, a 4 percent equipment increase is a few hundred dollars. If your CSR calls a customer and screams about a price deadline over $300, that customer will price-shop you out of spite, and they should. The honest argument is stronger anyway. It has two parts: the second failure, and availability.
The second failure is the real cost. A homeowner who defers again pays for the next emergency repair, then still buys the system, and pays next year’s price for it. Add up the repair spend already on the account and show the math. The availability argument comes from Carrier’s own disclosure that residential field inventory ended Q2 about 25 percent below the prior year. Thin channel inventory in a hot season means the scheduled August replacement goes in next week, and the January emergency replacement goes in whenever a matching unit shows up. Operators who sell that difference honestly do not need a fake deadline.
The decision to make before September 1
Pick one thing and commit for 30 days. Either somebody in your office owns the deferred-replacement list starting Monday, with a named person, a weekly call target, and a number you check on Friday, or you skip it entirely this season and stop pretending it is on the roadmap. A half-run list produces worse results than no list, because your best customers get one awkward call and never hear from you again.
If you commit, keep it small. Top 40 accounts by score. One CSR, ten calls a day, four days. The offer is a free replacement assessment with a written good-better-best number, not a sales visit. Track two numbers only: assessments booked and assessments sold. Anything below a 30 percent booking rate on this list means the script is wrong, not the list.
The signal to watch over the next 30 days is Q3 manufacturer commentary and whether field inventory keeps falling. If Carrier and Lennox both report replacement volume improving into Q4 while inventory stays thin, the shops that built their list in August will be quoting from a position of scarcity. The shops that waited will be competing on price in a market where their own costs went up 4 to 7 percent. That gap is entirely a function of work you can do this week with a CSV export and an afternoon.
At Atlas Unchained we build the boring operational plumbing that turns data you already own into booked revenue, for trades businesses that would rather run the shop than run software. If this is the kind of thing you want landing in your inbox every weekday morning, subscribe below.
