How do you choose an answering service for a roofing company?

You're in a driveway with a tape measure in one hand and your phone in the other, and a sales rep is telling you their answering service handles roofing. Of course it does. Ask whether they answer around the clock, or take a message, or handle storm volume, and the answer is yes every time. Nobody on a sales call is ever going to tell you no, which means the answers you get decide almost nothing and the tests you run decide everything.

Roofing makes a bad pick expensive in a specific way, because your phone doesn't fail on an average Tuesday. It fails the week a hail line comes through overnight and the calls stack up faster than anyone in the office can work them, then again when the door-knockers hit the same street. So shop it the way you'd inspect a roof: get up there and look at the parts nobody photographs for the brochure.

The seven checks, in order

Choosing an answering service for a roofing company comes down to seven checks, and you run every one of them yourself instead of asking about them on a sales call. If it is an AI receptionist you are weighing rather than a human service, the checks a roofer should run are slightly different. In order:

• Check 1. Pull a week of your own call data, so you know which hours you're buying.
• Check 2. Call them yourself at a bad hour, and listen for whether they ask for the job.
• Check 3. Make several calls land in the same minute, and hear what the third caller gets.
• Check 4. Read the billing unit before the brochure.
• Check 5. Follow one call past the message, to your calendar.
• Check 6. Get a name for 11pm on a Sunday, not a policy.
• Check 7. Score booked inspections, not calls answered.

Check 1. Pull your own numbers before you shop anything

Most owners know they miss calls and nothing past that. Pull the log off your business line and sort a week four ways: total inbound, how many got answered, how many landed outside normal hours, and how many came in on a weekend. If your number forwards through a cell or a tracking number, that dashboard is where this lives. Pick the right week too, because your average month isn't what you're buying coverage for. Take the week after the last wind event that put you on ladders.

Borrow one move from the research while you're sorting. Invoca's 2026 home services benchmarks report the answer rate two ways, 65% for calls that ran past 15 seconds and 73% for calls past 30 seconds, a cut they describe as stripping out misdials and quick hangups. Run the same cut on your own log, because a three-second call isn't a lost job and counting it as one has you buying coverage for a gap that isn't there.

What you end up with isn't "we should get answering," it's a brief: the specific hours your own log says you're losing. Where that gap usually sits for a roofer is broken down in why roofers miss so many calls during storm season.

Check 2. Call them yourself, as a homeowner, at a bad hour

Don't ask what their intake sounds like. Call the number they give you and be a homeowner, twice: once around 7:40 on a Saturday, once around nine on a weeknight. Those are the hours a scheduled demo never covers.

Listen for four things. How long before somebody's actually there. Whether a menu comes first. Whether they ask what a roofer would ask: is water coming in right now, when did the storm hit, is a claim open, what's the address. And whether they can tell apart the people who share your line, because a homeowner, an adjuster, a supplier and your own subs all dial the same number. A trade-built intake sounds like a fixed list of fields rather than a friendly chat, and the same intake written out for an HVAC no-heat call shows you what you're listening for.

Then the one almost everybody skips: do they ask for the job? It's a common miss, not a rare one. Per that same Invoca report, 55% of home services businesses don't ask leads to buy or book. That's measured on companies handling their own phones rather than on answering services, but it's the same miss to listen for when somebody handles yours. Then ask what you'll get asked on real calls: is this covered by my insurance? You want them to capture that and hand it to you, not answer it.

Check 3. Make the phone ring three times at once

This is the roofing-specific one, and almost nobody thinks to run it. Get two or three people to dial the number inside the same minute while you sit on one line. What you're buying isn't what the first caller hears, it's what the third caller hears, because the morning after a wind event the third caller is the normal case. Either they get answered like the first, or they get hold music, a busy tone, or voicemail. Run it again on your real forwarded number once a trial starts, since a sales line isn't the queue you'd be buying.

The human-staffed version deserves the fair read here. A live answering service puts a trained person on the line, which is genuinely worth something on a complicated call. But those agents are a pool shared across clients, and after hail a lot of roofers in the county are routing overflow into the same one. Ask what happens when their floor is slammed, and what the ceiling is. A straight answer is a good sign. "That doesn't happen" is not.

Check 4. Read the billing unit before you read the brochure

Pricing structure matters more in roofing than in almost any trade, because your volume isn't a line, it's a series of spikes. The month you most need the coverage is the month the bill runs highest, and payroll and materials are peaking in that same month.

So before the feature list, get the billing unit in writing plus four answers. Per minute, per call, or flat. That first split, whether the meter runs by the minute or the price sits flat, is what decides your bill in a storm month. If per minute, is it rounded up. How are automated, dead-air, and wrong-number calls counted. And are existing customers billed the same as a new lead.

That third one isn't hypothetical. Unwanted calls are a measured national problem: the FTC logged more than 2.6 million Do Not Call complaints in fiscal year 2025. How much reaches your line only your own log from Check 1 can say, and on a per-minute plan every one of those is billable time. So hand them that week and ask them to price it.

Check 5. Follow one call all the way past the message

A service picking up the phone isn't the product. What happens in the next four minutes is. So hang up and start a timer. Where does the message go: email, text, a portal? Does it reach the roofing CRM you already run, or are you retyping every booking? And does anything land on your calendar, or in a list somebody still has to work?

That distinction decides most of it. A message is a to-do item you now own. A booked inspection is a job with a time on it. The homeowner's clock doesn't stop when somebody picks up, it stops when there's an appointment on the books, and every minute in between is one a competitor can book instead. That gap is the subject of how an AI receptionist and an answering service compare for contractors. So the standard isn't "do you take messages accurately." It's "what's on my calendar when this call ends."

Check 6. Get a name for 11pm on a Sunday, not a policy

Ask what happens at 11pm on a Sunday and you'll get a sentence: after-hours calls are escalated per your instructions. That means nothing. Push until you get a mechanism. Who decides a call is worth waking you for, and against whose definition of urgent?

Write that definition yourself: water inside is obvious, a blow-off with rain in the forecast is too, a two-year-old ceiling stain isn't. If you want a starting point to argue with, one worked version of a roofing urgency line splits it into emergency, urgent, booked, and the callers who get qualified first. Then ask where the escalation lands, and what happens if that person doesn't pick up.

Here's the part specific to your trade. The escalation path most services quietly assume, transferring the caller straight to the owner, only works when the owner can pick up. Plenty of days you can. The morning after a storm, in an adjuster meeting, or up a ladder, you can't, and those are the hours you're buying.

Test it on your trial: put a call in around nine at night, say up front that it's a test so nobody rolls out, and watch what reaches you. And don't let the fallback be a voicemail box, because most voicemail-bound callers never get a call back, as covered in whether customers actually leave a voicemail when you don't answer.

Check 7. Run two weeks, and score booked inspections

Every service reports calls answered. It's a big number and the wrong one. Across home services, 38% of calls answered by a person are leads, and 45% of those leads convert on the call, per Invoca's 2026 benchmarks. Most of what gets answered was never a job, so "we handled 214 calls" tells you nothing. Roofing doesn't convert on the call the way a same-day trade does either. Your scoreboard is the booked inspection.

Score these four instead:

• Booked inspections with an address and a time on your calendar.
• Callers who asked for a person and got one.
• Calls you had to redo, because the intake missed something and you rang back yourself.
• Escalations that reached you, and how long each took.

Then pull call recordings at random, not the ones they send you, and listen all the way through. If there are none to pull, you can't run this check at all, which is itself the answer.

Where this checklist breaks

Two weeks in a flat month tells you almost nothing about a storm week, so start a trial heading into your season. If you can't, lean on Check 3, because a load test is the only proxy you'll get. And a test call is you, so you'll be clearer and more patient than a homeowner under a drip with a bucket. Get somebody who doesn't work for you to make one of the calls.

There's also a floor no service clears. If your first available inspection is a week and a half out, answering faster only gets you a faster no. Phone coverage fixes the calls you never got to. It doesn't fix a calendar.

Running the same checks on Willison

It's only fair to run our own list against us, including the parts we don't win. Willison answers every inbound call, 24/7, 365, in seconds, effectively on the first ring, and it answers many at once, so the tenth caller on a storm morning gets answered as fast as the first, with no busy signal and nobody on hold. That's Check 3, and Check 6 has the same answer: at 11pm on a Sunday the thing that picks up is Willison, and it can text you the details when a call needs you rather than routing to a name who may not answer.

It works out what the job is and books qualified work straight to your calendar instead of handing you a message, which is Check 5.

Check 7 has something behind it too: every call runs on a recorded line, gets transcribed, and is kept for 395 days. That is what lets you score booked inspections instead of calls answered.

Where it stops, plainly. Willison doesn't get on the roof and it doesn't close the job. It doesn't dispatch a crew, send a truck, or promise an arrival time, and it won't tell a homeowner what their carrier will cover. It doesn't fix a booked-out calendar either. And Check 4 is the one no blog post should answer, ours included: make somebody put the billing unit in writing before you sign, us included.

Check 2 you can start right now. Talk to the live Willison demo right in your browser on willisonhq.com and put a real roofing call to it. It won't reproduce a Saturday at 7:40, so that test still stands, but it'll tell you in ninety seconds whether the intake is worth testing at all.

Frequently asked questions

That varies by service, and it's the biggest real difference between two options that look identical on a website. Message-taking hands you a to-do item you still have to work. Booking puts an address and a time on your calendar while the caller is on the line. Ask to see what lands on your side the moment a call ends.

It depends on how many calls the setup holds at once, which for a human service means agents on shift rather than phone lines. When the floor is full the extra callers queue, and a homeowner watching water come through a ceiling doesn't wait in a queue. They dial the next roofer on the page.

Two weeks is plenty for judging intake quality, the handoff, and the escalation path, since those behave the same every day. It isn't enough for peak load, because a quiet fortnight never puts pressure on it. Start heading into your busy season where you can, and run a simultaneous-call test where you can't.