Home Services: Every Missed Hello Is a Job Ticket

In home services, the call is not a touchpoint. The call is the job. When a homeowner's air conditioner dies in July, they call three companies, and the first live human they reach gets the work. Not the best website. Not the best reviews. The first person who answers.

Now put that reality next to this one: the default detection in most dialers drops an estimated 10 to 20 percent of live human answers as if they were voicemail. That is AMDY's measured figure across the floors we have instrumented, and it applies to home services the same as any vertical. But in no other vertical does the dropped call convert to a lost job ticket as fast as it does here. A missed hello in July, in a heat wave, with a customer holding a dead thermostat, is not a lost touch. It is a competitor's install, signed before your dialer finishes its next pass.

I ran outbound floors for years before building detection tooling, and home services was always the account type where the economics were the most brutal and the least understood. This article is the conversation I had with those owners, minus the jargon. It is about why your season makes every conversation precious, why your answerers are harder for stock detectors than anyone else's, and how to put a number on what you are currently losing, using your own figures, in one sitting.

Seasonal Demand Does Not Negotiate

Most businesses can absorb a lost conversation by spreading pain across the year. Home services cannot, and the reason is the shape of the demand curve.

An HVAC company earns a large share of its revenue in a handful of weeks: the first sustained heat, the first real cold snap, the storm that rolls through and takes compressors with it. A roofer's year is measured in hail seasons and wind events. A plumber's peaks ride frozen pipes and holiday weekends. You staff up for those windows, you pay premiums for labor and leads during them, and then the window closes until next year.

Here is what that does to the value of a single answered call. In a flat-demand business, a dropped conversation is a delay; the prospect is still reachable next week. In peak season, the dropped conversation is final. The homeowner with water coming through the ceiling is not waiting for your second dial attempt. They are answering the next company on their list, and the research on how consumers behave when a call goes unanswered is not kind: the Federal Trade Commission's own consumer guidance on robocalls and unwanted calls exists precisely because silent calls are among the most complained-about call experiences in America. Silence reads as a scam. Silence reads as a company that does not have its act together. Either way, the homeowner moves on, and they move on within minutes.

So the arithmetic that owners do in flat businesses, "we lost a touch, we will touch again," is exactly wrong for yours. The correct arithmetic is: we lost a job ticket, and the season will give us a finite number of chances to replace it.

The Peak-Week Math, In the Owner's Terms

Let me make that concrete with a formula rather than a case study, because I do not believe in inventing statistics about your business. You know three numbers off the top of your head: your average ticket, your close rate on reached humans, and your seasonal call volume.

Write this down with your own figures:

Lost revenue per week = weekly answered calls × drop rate (use 0.10 and 0.20 as the range) × close rate on reached humans × average job ticket

Run it twice, once with each end of the drop range, and you get the honest bracket. For an owner who never wants to overstate the case, use the low end. The point of the exercise is not to produce a scary number for a board deck. It is to find out whether the leak is worth five minutes of attention or a quarter of planning. I have watched owners run this on a napkin and go quiet, and I have watched a few run it and shrug. Both reactions are legitimate. The formula's job is to tell you which one is yours.

Notice what the formula does not include: agent wages saved by dropping calls. That is deliberate. In peak season, home services agents are not a scarce resource to be shielded from voicemail. They are idle capacity waiting for conversations, and your bottleneck is reached humans, not agent minutes. Protecting agent time by dropping live answers in your busy season is like a restaurant turning away tables to keep the waitstaff fresh.

Why Home-Services Answerers Break Stock Detectors

Here is the technical part of this article, and I will keep it in plain English, because the money part depends on it.

Answering machine detection listens to the first moment of a call and decides: human or machine. Stock detection does this with old pattern rules, and those rules assume a clean world. The greeting is either a crisp recorded message or a clear live voice in a quiet room. On that assumption, the rules were reasonable. Your call traffic violates every clause of that assumption, in three specific ways.

The Gatekeeper Problem

The person who answers a homeowner's phone is not always the homeowner. It is a spouse, an adult child, a property manager, an office receptionist at a small landlord company. Gatekeepers answer differently than prospects. Short greetings. Flat delivery. "Hello" and nothing else, because they answer calls all day and their voice has no novelty in it.

To a stock detector, a clipped, flat "hello" sounds closer to a voicemail prompt than to the warm pickup the rules were tuned on. The detector guesses machine. The call drops. The spouse of a homeowner with a failing furnace hangs up on silence, and the story in their head is that your company called and had nobody on the line. In trades where reputation is half the marketing budget, that story does damage beyond the lost ticket.

The Noise Problem

Home services means mobile phones, and mobile phones mean the world's worst recording environments. Roofer called on a job site: wind hammering the mic. HVAC tech driving between calls: cab noise. Contractor standing next to a running air handler: constant mechanical roar. Every one of these environments shreds the audio features a stock detector depends on, and shredded features push the guess toward machine, because machines sound like noise to rules written for clean lines.

The Carrier Greeting Problem

The third one is not your fault and not your traffic's fault, but you pay for it anyway. Modern carrier voicemail greetings are often warm, conversational, human-recorded messages, sometimes with a person's name and a casual delivery. They were designed to sound friendly. The side effect is that they sound alive, and the stock rules, tuned when voicemail meant beeps and robotic prompts, cannot hold the line between a friendly voicemail and a hesitant human. When the rules get unsure, they err in the most expensive direction: they drop the call.

We cover that asymmetry in depth here: https://amdy.io/blog/amd-error-directions. The short version for your whiteboard: a dropped human costs you the whole call plus the job it might have booked; a voicemail passed to an agent costs you some seconds of payroll. One is a revenue event. The other is an efficiency event. Any detection decision that trades the first to avoid the second is a bad trade in your vertical, in your season.

The Value Math, With Your Inputs

Let us build the full per-call model, because owners deserve to see the whole shape of it, not just the seasonal snapshot. Three costs stack on every dropped live answer.

Layer One: The Job You Did Not Book

The first and largest cost. A reached, live, interested homeowner had a need and a pulse and your company on their screen. The formula from earlier applies per call, and the unit version is simpler:

Value of a reached human = close rate on reached humans × average job ticket

Your close rate on reached humans is the number to be honest about. Not your close rate on appointments, not on leads. On conversations with live people who picked up. Most owners can estimate it well enough from last season's numbers, and the estimate is plenty for this decision.

Layer Two: The Season Multiplier

In home services, the value of a reached human is not constant across the year. In peak weeks it is higher, because the same conversation has a better close rate (urgent need) and a bigger average ticket (emergency pricing, full-system replacements). This is why I tell home services owners to run the leak math twice: once with shoulder-season inputs, once with peak inputs. If the leak is tolerable in April and intolerable in July, then the fix is a seasonal decision, and you time it before the season, not during.

Layer Three: The Reputation Ratchet

The third cost is the one nobody puts in a spreadsheet, so I will just say it plainly. Dead air after "hello" is the signature experience of scam calls. Your dropped calls are training your market to associate your caller ID with that experience. In a business built on trust, kitchens, roofs, basements, that association has a cost. I cannot give you a number for it and neither can anyone else, and you should distrust anyone who does. But every owner I have talked to knows it is not zero.

The Vertical View: Where Home Services Sits

Owners often ask how their vertical compares. Rather than inventing comparative statistics, which I refuse to do, here is the structural comparison, using real, checkable facts about each vertical and our published plan structure:

Vertical	Why the dropped call hurts	What drives the value of one saved conversation
Home services (HVAC, roofing, plumbing)	Seasonal windows and urgent needs; the homeowner calls the next company within minutes	Average job ticket is high (four figures common on repairs, more on replacements) and peak-season demand is inflexible
Insurance (agency outbound)	Renewal and quoting conversations compete against direct writers calling the same lists	Retention economics; we cover the full picture here: https://amdy.io/blog/amd-insurance-outbound
Any outbound vertical under FTC telemarketing rules	Dead air from dropped calls counts toward the 3% daily abandonment cap, per the rule text at ecfr.gov	Compliance headroom; fewer dropped live answers means fewer abandonment events to manage
Every vertical, regardless of trade	The default dialer drop range of an estimated 10-20% applies before any vertical effect	The fix is priced per detection: Sandbox is free at 50,000 detections per month, and paid tiers start at $79 per month including 500,000 (https://amdy.io/pricing)

Read the last row twice, because it is the one that makes the decision easy. The pricing is per detection, not per seat or per server, and plans include a monthly volume with overage at a flat per-detection rate. Unlimited servers, monthly terms. A home services operator running a seasonal surge can scale through the busy months and back down after, without a contract renegotiation. Our sibling piece on the insurance vertical walks the same math for agencies, and the comparison is instructive: insurance has smaller tickets but year-round volume; you have bigger tickets compressed into windows. Different shapes, same leak.

Two Scenarios Every Owner Will Recognize

Let me put the three detection problems on the street, because abstractions do not move budgets and Tuesdays do.

The first scenario: the gatekeeper drop. A storm rolls through a suburb on a Thursday. Your roofer list for that neighborhood goes out Friday morning. A homeowner's wife answers, says "hello" once, flat, because she has been fielding contractor calls for eighteen hours. The stock detector reads the flat greeting as a machine and drops the call. She hangs up, redials the next name on her own list, and by the time your dialer cycles that number again the roof is contracted. Nobody at your company even knows the conversation was available. The call shows up in the report as an answering machine, which is the one thing it was not.

The second scenario: the slow verdict. A homeowner answers on the third ring, standing in a hot garage next to a failed condenser. The detector on the other end wants two more seconds of audio before it commits. Silence stretches. The homeowner says "hello" again, gets nothing, and his read is instant: robocall. He hangs up and blocks the number. The detector eventually returns a confident, correct verdict, human, delivered to a dead line. This is why decision speed belongs in the same breath as accuracy, and why we publish the 125ms figure: a verdict that arrives after the human has given up is not a verdict, it is an autopsy. The longer version of that argument, in money terms, is here: https://amdy.io/blog/amd-latency-vs-accuracy.

Both scenarios share a nasty property: your reporting cannot distinguish them from ordinary voicemails. The failure is invisible in the exact dashboards you use to run the business. That is why I keep pushing owners toward per-call logs rather than summary stats, and it is why the trial matters more than the datasheet in this category. We also watch for subtler failure modes on our side, like traffic patterns that resemble honeypot numbers planted to catch dialers, and there is a fuller writeup of that here: https://amdy.io/blog/amd-honeypot-detection.

What Fixing It Actually Takes

I want to close the loop on deployment, because owners rightly fear that any detection project means carrier changes, dialer migrations, and a consultant camped in your office for a month.

It does not, and you should hold any vendor to that standard. Our install is one command and takes about five minutes. No carrier change. It sits alongside your existing dialing and starts producing per-call verdicts immediately, with the first verdict arriving at 125 milliseconds from answer, one-eighth of a second, fast enough that the homeowner's first "hello" gets met with a live voice instead of silence.

And because I spent the first half of this article telling you to distrust unverified claims, apply that to me too. Do not take my word for the drop rate or the verdict speed. Run your own traffic and read your own logs. The Sandbox plan gives you 50,000 detections per month free, hard-capped, no card required, and per-call logging is on by default. You can count what your traffic actually does, on your lists, in your season. Signup is at https://amdy.io/auth/signup. If you want a longer look at the latency-versus-accuracy trade in money terms before you do, that piece is here: https://amdy.io/blog/amd-latency-vs-accuracy, and the full feature set is documented at https://amdy.io/features.

The Season Does Not Wait

Every home services owner I have worked with keeps a mental count of something: the calls that got away. The bid lost by an hour. The quote sent a day late. Detection is the one leak in that count that is purely mechanical, purely silent, and fixable before your next peak week.

You already know your average ticket and your close rate. You already know when your season starts. The only question left is whether the leak is big enough to matter, and the formula above will answer it in the time it takes to finish a coffee.

So here it is, plainly: how many job tickets is your dialer dropping this season, and when were you planning to check?