Explainer
How profitable is an HVAC business? Where the margin goes.
Every page for this question is a list of industry averages, and the most quoted one comes from a study nobody can find. This page says what the public figures actually say and where they come from, then does the thing an averages page cannot: walks the P&L of an illustrative shop, line by line, and shows which back-office lines eat the margin and by roughly how much.
No email required. No gate. The shop below is an illustration with round numbers and the arithmetic shown, so you can put yours in. Sources are named and linked, and their limits are stated.
01
What the averages say, and where they come from
Search this question and you will read, on most of the first page, that the median HVAC contractor nets 5.8 percent and the top quartile 13.2, citing a 2024 ACCA benchmarking study. I went looking for the study. So did someone more thorough, who found that no study by that name appears in ACCA’s catalogue, that ACCA’s most recent published data is older, and that the 5.8 figure circulates through articles citing other articles with no primary source at the bottom. Treat it as a rumour with decimals.
What can be traced is thinner and more honest about what it is. ServiceTitan’s published guidance, which is a vendor’s view and not a survey, puts a healthy residential shop at ten to twenty percent net, says many shops sit below ten and some at two or three, sets fifty to fifty-five percent gross as the target, and puts overhead at twenty to forty-five percent of revenue, thirty to forty-five for residential service. It ties all of it to one driver, the field’s billable utilisation, which is a fair thing to tie it to.
And one hard number from a public source: the Bureau of Labor Statistics puts the median wage for the trade at $61,010 a year, $29.33 an hour, as of May 2025, across 440,900 people. That is the one input below that is not mine.
So the honest answer to “how profitable” is a band, not a number, and the band is wide: a shop can net three or eighteen on the same revenue. What decides where it lands is mostly overhead, and the overhead a shop can actually move lives in the back office. That is the rest of this page.
02
An illustrative shop, line by line
A residential service and replacement shop: 8 techs, about 2,500 jobs a year, $2 million in revenue. The percentages are round and sit inside the public bands above. It is not a client. Change any line to yours and the rest recalculates in your head.
| Line | Share |
|---|---|
| Revenue | 100% |
| Equipment and material | 30% |
| Field labour, loaded | 22% |
| Gross margin, what the field produces | 48% |
| Office labour, three people, loaded | 9% |
| Marketing and lead fees | 7.5% |
| Owner’s salary | 6% |
| Vehicles, fuel, maintenance | 6% |
| Insurance, licences, bonding | 3% |
| Rent, utilities, yard | 2.5% |
| Software, phones, card fees | 1.5% |
| Everything else | 4.5% |
| Overhead, where the back office lives | 40% |
| Net, what is left | 8% |
Eight percent net on two million is $160,000, and it is the number the owner sees. Read the table the other way. The field produced $960,000 of gross margin, which is a good year. The back office and everything around it spent $800,000 of it before the owner saw the eight. That is not a criticism of the office; that is what overhead is. The question is which of those overhead lines are doing work and which are re-doing it.
The three lines the owner cannot easily move are equipment, insurance and rent. The one everybody tries to move is marketing. The lines nobody looks at are office labour, the collection lag hiding inside “everything else,” and the software row, because each is small. The next section is about those.
03
The back-office lines that eat the margin
Five, on the illustrative shop, each with its assumption stated and its arithmetic shown, so you can disagree with the assumption and redo the sum. Four are overhead and add up to points of net. The fifth is revenue, and it is bigger.
-
01
The re-typing
1.2 points of net
Seventeen hours a week of office time spent typing what another screen already knew, the count from the quote-to-invoice trace, at a loaded office rate of $28 an hour.
17 × 52 × $28 ≈ $24,800 a year
The fix, and what it costs
Configuration you already own: the price book loaded, quote to job to invoice inside the tool, the sync on. The worked example is on the quote-to-invoice page.
From quote to invoice without typing it twice → -
02
The part that never made the invoice
0.75 points of net
One job in ten leaves the site with something not billed: the expansion tank, the extra half hour, the second filter. Call it $60 a time on 2,500 jobs.
250 × $60 = $15,000 a year
The fix, and what it costs
The tech closes the job on the app before leaving the driveway, with the parts on it, and the invoice is built from the job rather than from a text the next morning.
Run the job, on the leaks map → -
03
The money that arrives late
0.7 points of net
Receivables sitting at forty-five days instead of twenty on $2M of revenue is roughly $140,000 more owed to you at any moment. If a line of credit is carrying it at ten percent, that is the cost of the wait.
$140,000 × 10% ≈ $14,000 a year
The fix, and what it costs
Invoice from the closed job the same day, card on file for service work, and an automated reminder at seven and fourteen days, which the tool you pay for already has.
Get paid, on the leaks map → -
04
The software used a fifth
0.6 points of net
A $30,000 a year stack of which the features actually used are worth $18,000, because the marketing suite, the memberships engine and the reporting were switched on once.
$30,000 − $18,000 = $12,000 a year
The fix, and what it costs
Cut to the package that covers what you run, at renewal, or configure the rest. Rarely a new product.
ServiceTitan alternatives: configure, cut, or move → -
05
The quote that went out on day three
Not an overhead line, a revenue one, so it is not in the points below. But a quote sent three days after the visit closes less often than one sent the same evening, and you know your own close rate. If it moves five points on $600,000 of quoted work, that is $30,000 of jobs, and the gross margin on them.
Your close rate, your quoted volume
The fix, and what it costs
The price book, again. A quote assembled from a loaded book goes out from the driveway; one assembled from a supplier PDF goes out on Thursday.
Win the work, on the leaks map →
The sum, on this shop
The four overhead leaks come to about 3.25 points of net: from 8 to roughly 11.3 percent on the same revenue, with the same techs, the same trucks and the same marketing spend. That is the distance between the “average” and “healthy” bands in the public guidance, and none of it needed a hire or a new tool. The fifth leak, the late quote, sits on top of that in revenue, and it is the one the calculator asks you about.
04
Admin cost per job, the sum nobody does
Every shop knows its material cost per job and its labour cost per job, because the estimate has lines for them. Almost none knows its office cost per job, because the office is a lump called overhead and nobody divides it. Divide it.
On the illustrative shop: three office people at forty hours is a hundred and twenty loaded hours a week, across roughly forty-eight jobs a week, at $28 an hour. That is about $70 of office time on every job, of which the re-typing alone is about $10. On an $800 average ticket, $70 is nearly nine percent of the revenue of the job going to the desk before the truck has moved. It is a real cost and most of it is real work. The part to look at is the $10, because it is the part that is the same job being done twice.
Do it for yours. Office hours in a week, divided by jobs closed that week, times the loaded rate from payroll. Ten minutes. Then ask the office manager, honestly, what share of the week is typing something that was already typed somewhere else. That share of the number is the leak, and it is the number the audit puts beside each arrow.
05
Put your own number on it
The calculator on this site asks four things: the hours a week the office spends on the manual process, the loaded hourly cost, the error and rework rate, and whether it slows or loses customers. It gives back a yearly figure and a payback. It is rough by design, and it is the same arithmetic as the section above with your inputs instead of mine.
The exact version, with the loaded rates from your payroll, the re-types counted with the people who do them across every job type, and a dollar figure beside each leak that the owner will believe, is the AI Audit. One job traced through the shop, the leaks priced, a written blueprint ranked by what each recovers, and for each one: configure, connect, build, or leave it alone.
A fixed fee, $5,000, sized by headcount. If the recoverable savings in the blueprint do not cover it, you do not pay, and you keep the blueprint. On the illustrative shop the four overhead leaks alone cover it several times in the first year; on yours, the count will say.
06
Questions people actually ask
How profitable is an HVAC business?
Less than the averages suggest and more than most shops achieve. The public guidance that can actually be traced says a healthy residential shop nets ten to twenty percent, many sit under ten, and some run at two or three. The widely quoted “median of 5.8%” comes from a study nobody can find. What decides where a given shop lands is mostly overhead, and most of the overhead that is controllable lives in the back office.
What is a good profit margin for an HVAC company?
On the vendor guidance that names its figures, fifty to fifty-five percent gross across services and ten to twenty percent net, with the field’s billable utilisation as the strongest single driver of both. Treat those as targets, not as a survey. A shop netting eight on two million revenue is not failing; it is carrying three or four points of leak it could recover without hiring anyone.
What is overhead in an HVAC business?
Everything that is not equipment, material and the field’s wages: office labour, marketing, the owner’s salary, vehicles, insurance, rent, software, card fees. Public guidance puts it at twenty to forty-five percent of revenue, higher for residential service. Overhead is where the back office lives, and it is where the margin goes when the same job is typed four times.
How much does admin cost per job?
Divide the office’s loaded hours in a week by the jobs completed in that week and multiply by the loaded rate. On the illustrative shop, three office people at forty hours is a hundred and twenty hours a week across about forty-eight jobs, at $28 an hour: roughly $70 of office time per job, of which the re-typing alone is about $10. Do the same sum for your shop; it takes ten minutes and it is the number nobody has.
How do I increase the profit margin of my HVAC business?
Before pricing or marketing, close the leaks that cost nothing to close: load the price book so quotes go out the same day and convert to jobs with lines on them; make techs close on the app; invoice from the closed job; switch on the reminders and the sync; cut the software you do not use. On the illustrative shop that is three to four points of net, which is the difference between the “average” and “healthy” bands in the public guidance, and none of it needed a new hire or a new tool.
Can you tell me my shop’s number?
The calculator will give you a rough one from four inputs. The exact one, with the loaded rates from your payroll and the re-types counted with the people who do them, is the AI Audit: one job traced through the shop, the leaks priced, a written blueprint ranked by what each recovers. A fixed fee, $5,000, sized by headcount, not owed if the savings do not cover it.
Sources, checked 18 September 2026
- U.S. Bureau of Labor Statistics, Occupational Outlook Handbook: HVACR mechanics and installers Median wage $61,010 a year, $29.33 an hour, May 2025. 440,900 employed in 2025.
- ServiceTitan, “Understanding HVAC profit margins” A vendor’s guidance, not a survey: 10–20% net is “healthy,” 50–55% gross is the target, overhead runs 20–45% of revenue (30–45% for residential service), and many shops sit below 10% net, some at 2–3%.
- VWCG, “The HVAC benchmark that does not exist” Traces the widely quoted “2024 ACCA benchmarking study” (5.8% median, 13.2% top quartile) and finds no such study in ACCA’s catalogue; the figure circulates through articles citing articles.
Related