We called 20 HVAC companies after hours. One could help.

We called 20 US HVAC companies during their own 6 to 8pm window, across four time zones and several evenings, posing as a homeowner with an air conditioning problem.

Of the five where a human picked up, one could actually help that evening. We left 7 voicemails with a working callback number. None were returned within 48 hours. The only two companies that called back were ones where no message was left at all. They saw a missed call and chased it.

Why we ran this

Everyone in home services marketing says the same sentence: you’re losing money to missed calls. It’s repeated so often that owners have started treating it as a sales line rather than a fact, and fairly so, because almost nobody says where the number comes from.

We wanted to know what actually happens. Not what a survey says owners believe happens on their phones. What a customer experiences when they dial at 7pm on a weeknight. So we called, and wrote down what happened.

Method

Sample: 20 residential HVAC companies in the United States, each an independent small business of 1 to 10 employees, spread across Eastern, Central, Mountain and Pacific time zones.

Timing: each company was called once, during their local 6:00 to 8:00pm window, across several different evenings so that one unusual night could not distort the result. The first ten calls were made on a single morning; the remaining ten over the following mornings.

The call: the caller presented as an ordinary homeowner with an air conditioning problem. If a person answered, the question was simple, do you take service calls in the evening, or should I call back tomorrow?, followed by a polite exit. No appointments were booked and no company’s time was wasted beyond the length of the call.

What was logged: rings before answer, who or what answered, whether the greeting promised a callback and in what timeframe, whether a message could be left, and the exact time any message was left.

The callback test: where voicemail allowed it, one message was left with a name, a brief description of the problem, and a working callback number. Every message was then watched for 48 hours.

What answered the phone

Answering the phone is not the same as capturing the customer

This was the result we didn’t expect.

Five companies had a real person pick up. Only one of those five could actually help that evening, a Florida company where the person asked what service was needed and confirmed they handle after-hours calls.

The other four:

• One answered, and when asked about evening service, said to call back tomorrow. No details taken, no callback offered.
• One rang for roughly a full minute before a person answered, then explained it was after hours and would be handled tomorrow. A homeowner with water on the floor would have hung up long before that.
• One assumed the call was about commercial work, said they don’t do commercial, and ended the call. The word “commercial” was never used by the caller. A residential customer was lost in under ten seconds.
• One answered and the call dropped mid-sentence. Nobody called back. From the customer’s side, that is identical to not being answered at all.

Two of the twenty calls reached somebody who could act: the Florida company, and one AI agent that heard the problem and transferred the caller to a live person. That is a 1 in 10 chance that a homeowner with a failing system on a weeknight evening reaches help.

Most conversations about missed calls stop at “did the phone get answered”. This study suggests that’s the wrong question.

The callback promise is mostly decoration

Six of the greetings promised a callback in some form, two with specific hours (“we’re open 9 to 5”) and four vague (“your business is important to us, we’ll get back to you”).

Seven voicemails were left with a name, a problem, and a working number. None were returned within 48 hours. Not late. Not partially. None.

We want to be careful here: a sample of seven messages is small, our number was unfamiliar to them, and any one company might have been having an unusual week. But seven out of seven is still seven out of seven, and every one of those messages was a person asking to give them money.

The businesses that called back were watching missed calls, not reading messages

Two companies did call back.

Neither had received a message. In both cases the call simply went unanswered or ended, and somebody later looked at the caller ID and dialled back.

That is worth sitting with. The mechanism that recovered the customer wasn’t the voicemail system. It was a business with the habit of noticing that someone had called and not reached them.

One of those two was also the best-handled call of the entire study. An AI assistant answered, asked what was going on, and the moment the caller said their AC was acting up, it transferred them to a live person. Same company, twice: a system that caught the call, and a habit that caught the miss.

AI is already answering these phones

Three of the twenty companies, 15% of a random sample of small independent HVAC shops, answered with an AI voice agent. This is no longer a coming trend in this industry. It’s in the field now.

But the quality gap was immediately visible:

• The good one heard a real problem and put the caller through to a human.
• The other two were polite, natural-sounding, and ultimately ended the call by telling the customer to call back during business hours. One even mentioned that after-hours emergency service existed, then directed the caller to phone back the next day.

Answering pleasantly and losing the job is still losing the job. Two of the three AI agents produced exactly the same outcome as voicemail, delivered in a nicer voice.

What we’d tell an owner reading this

Nothing in this study was about technology. Four things separate the companies that captured the caller from the ones that didn’t.

1. Somebody or something answers within a few rings. A minute of ringing is a lost customer regardless of who picks up at the end of it.
2. Whoever answers can take the next step. Get the details, offer a time, or reach a person who can. “Call back tomorrow” is an outcome, not a courtesy.
3. Nobody assumes what the caller wants. The company that decided the call was commercial lost a residential job without ever asking a question.
4. Somebody looks at the missed calls. It was the only thing that recovered a customer in this entire study.

Caveats, stated plainly

• 20 companies, one call each. This is directional, not statistically definitive.
• Every company was called once, in one window. A different evening might have produced a different result for any individual business.
• Our callback number was unfamiliar and originated outside the United States, which may have affected how some companies treated it.
• One callback arrived while the line was unattended, so it is counted as a callback attempt rather than a completed conversation.
• We are not neutral. Botspire builds AI phone systems for HVAC and plumbing companies. We ran this study because we’d rather know what actually happens than repeat an industry talking point. The results were not adjusted to flatter that position, including the finding that two of the three AI agents in the sample performed no better than voicemail, and that the single best call of the study belonged to a company that is not our customer.
• No company is named in a negative context in this write-up, and no recordings were made.

What we’d do differently next time

Larger sample, plumbing companies included, a second call to the same businesses during peak season, and a wider callback window. If you’d like to see that study when it exists, or you run a shop and want to know what happened when we called your line, get in touch.

Botspire builds AI phone systems for HVAC and plumbing companies: answering, booking, emergency routing, and 60-second callbacks on web leads.