HVAC after-hours calls: why 62% go unanswered — and what each miss actually costs.

Sunday night, 9:47 PM, upstate New York, first hard freeze of October. Furnace in a three-bedroom on the east side stops. Homeowner Googles "HVAC emergency near me." Three shops show up. She calls the first one — voicemail. Calls the second — an answering service somewhere in Nebraska that says "we'll have someone call you back within 4 hours." She calls the third. They pick up on the second ring, dispatch someone by 10:20, invoice on Monday: $487 emergency after-hours service plus a $340 same-day capacitor replacement.

The first shop she called was ours — before we fixed it. They lost that job at 9:47:04 PM, the second the voicemail beep played. And that specific pattern repeats about 40 times a month across upstate New York's HVAC operators. I've done the math on eleven shops now. It's always uglier than the owner thinks.

The number nobody wants to look at

When we onboard an HVAC operator we pull 60 days of their Twilio, Google Voice, or landline call logs and cross-reference against dispatch records. Here's the pattern that shows up in every single shop under 12 techs:

Multiply through: 40 after-hours calls a month × 62% unanswered × 8% "recovery" via callback × $340 average ticket. The unrecovered value — calls that permanently walked to another shop — is roughly $8,000 per month per shop. That's not a marketing miss. That's a working-capital hole.

Why answering services don't work

The answering-service industry sells a specific promise: "we'll take the call so you don't have to." That promise made sense when the alternative was voicemail. It stopped making sense the moment the homeowner had four other tabs open on the same search page.

The problem isn't that answering services are lazy. It's that they can't actually solve the emergency. They log a ticket. They page a tech. The tech calls back in 20 minutes. By then the homeowner is on the phone with the shop two zip codes over that answered on the second ring. The answering service does the right thing per their playbook — and the homeowner is already gone.

The other issue is qualification. A homeowner at 10 PM saying "my furnace is out" needs three questions answered before dispatch: is it gas or oil, is there a burning smell, is anyone in the house medically compromised. An answering service reads a script. It doesn't know that a burning smell means we're calling National Grid instead of dispatching, or that a CO alarm means 911 goes first.

What actually works

Two things, both cheap, both non-obvious:

1. Answer in 45 seconds, always. The gap between the third ring and voicemail is where the customer decides you're closed. Whatever answers has to sound human, know the local zip codes, ask the three triage questions, and either dispatch or book for tomorrow morning. Whether that's a human, an AI, or a hybrid isn't what matters. What matters is that at 9:47 PM the phone gets picked up.
2. Trigger a text within 90 seconds of any missed call. "Hi — this is Rachel with Northern HVAC. I see you called, our team is on another emergency, but I can dispatch someone in the next 30 minutes if you can text me what's going on." That single SMS recovers about 40% of the calls the answering service would have lost, because the homeowner hasn't closed the tab yet.

The math on the two together: a shop doing 40 after-hours calls a month recovers roughly 22 additional dispatches per month. At $340 average ticket, that's $7,480/month in recovered revenue. On our fee, ROI is not the question — the question is why anyone still runs an answering service in 2026.

The three triage questions that actually matter

Whatever you deploy — an AI, an in-house night receptionist, or a better-trained answering service — the same three questions get asked before dispatch:

1. Gas, oil, or heat pump? This determines who dispatches and who's called. A gas leak isn't a service call — it's National Grid plus 911. Get this wrong and you're liable.
2. Any burning smell, CO alarm, or water on the floor? Any yes here shortcuts the whole flow to a safety protocol. A well-designed AI script asks this before it asks for the address, because "your address is X and please evacuate the house right now" is a different call than "we'll be there in 30 minutes."
3. Age of the system? Determines whether we're pricing a repair (under 12 years) or a service call that ends in a system quote (12+ years). Both are legitimate. But they route to different techs and different scripts.

Getting these three right is the whole difference between an emergency call handled well and one that ends in either a lawsuit or a review that reads "they didn't take me seriously."

Every unanswered emergency call is worth about $340 walking to your competitor. Every one you catch is worth $340 you didn't have to spend Google Ads dollars on. The rate of return on picking up the phone at 9:47 PM is roughly infinite.

What we don't recommend

We don't recommend paying your senior techs to be on-call rotation for the phones. They're worth $95/hour on a job. Not $12/hour reading a script. If they're on rotation, cap it at dispatch approval only — the phone gets answered by the AI or a night receptionist, escalates the qualified emergencies, and the tech says "yes go" or "book for 7 AM."

We don't recommend a chatbot on the website that says "we'll get back to you soon." At 9:47 PM the homeowner needs a person or a very good AI on the phone, not a form. If your website doesn't have a "call now" button as its dominant element on mobile, that's the first fix — before any AI conversation.

We don't recommend voice AI without a warm-transfer handoff to a real dispatcher. The AI qualifies, gathers the triage answers, and hands to your on-call in 20 seconds. It does not close the job on its own. That's the ethical line and it's also what customers actually want.

The point

62% of your after-hours calls are going to voicemail. That's the number. It's not because your team is bad — it's because 6 PM to 8 AM is 14 hours of coverage that costs $200/night for a person and produces value only when the phone rings. AI closes that gap for a fraction, but only if it's set up to actually answer, actually triage, and actually dispatch. Half-measures — chatbots, answering services, "we'll call you back" scripts — are worse than the voicemail they replaced.

The audit is free. I pull your last 60 days of call logs, tag every after-hours miss, and give you a spreadsheet with the specific dollar value walked. Then you decide whether it's worth fixing.