The Five-Minute Callback: A Speed-to-Lead Playbook for Roofing Companies

Most roofing companies that think they have a lead problem actually have a response problem. The leads are arriving. They are being answered by somebody else first.

This is the playbook for the fix: the complete system that puts a briefed human on the phone with a new roofing lead in under five minutes, during business hours and at 9 p.m. on a storm night. It is written for an owner running two or more crews who is already paying for leads, marketplaces, or ads, and watching too few of them become booked inspections.

First, a definition, because the whole system hangs on it. The standard is time to first human contact: the minutes between a homeowner's enquiry arriving and a person from your company speaking with them, or reaching them with a real attempt they can answer. An automated text does not stop this clock. Automation's job is to hold the homeowner and arm the human, not to impersonate one. Track the two numbers separately from day one, or the system will report success while the job goes to a competitor.

Build the stages in order. Each one removes a specific leak, and the leaks are cheaper to find in your own timestamps than in a lost month.

Why are my roofing leads not converting?

Start with the diagnosis most owners resist. The lead source is the first suspect because it is the only suspect that is not the owner's fault. So the vendor gets switched, the same conversion rate shows up a month later, and the variable that actually decides the outcome, the minutes between enquiry and first contact, never gets measured.

The published research on that variable is old, consistent, and stronger than the folklore built on top of it:

• The Lead Response Management study (Oldroyd and InsideSales, 2007) analysed three years of data, more than 15,000 leads and 100,000 call attempts, and found the odds of contacting a lead called within five minutes versus thirty minutes drop 100 times, and the odds of qualifying it drop 21 times. It measured contact and qualification, not signed contracts, and its data predates the smartphone. Later studies keep finding the same cliff.
• The Harvard Business Review audit (Oldroyd, McElheran and Elkington,
  2011. submitted test leads to 2,241 U.S. companies. The average first response took 42 hours among companies that responded within thirty days, and 23 percent never responded at all. In the article's companion dataset of 1.25 million leads, firms that tried to contact a prospect within an hour were nearly seven times as likely to qualify the lead as firms that waited even an hour longer, and more than sixty times as likely as firms that waited a day or more.
• A 2021 analysis of 5.7 million inbound leads across more than 400 companies found conversion rates eight times higher when the first call came within five minutes rather than after six or more. Under one percent of first attempts happened that fast, and 57 percent happened more than a week after the lead arrived (XANT, formerly InsideSales, 2021).
• Nothing about this has fixed itself. When researchers submitted real enquiry forms to 433 B2B companies, only seven percent responded within five minutes and 55 percent had not responded within five business days (Drift, 2017). A 2024 test of 1,000 B2B demo forms found 635 companies never replied at all (RevenueHero, 2024).

Roofing sharpens every one of those numbers. Verisk's 2026 U.S. Roof Report puts the average residential roof replacement at $17,631 in 2025, so a single leaked job is a five-figure event, not a rounding error. Roofing Contractor's 2026 homeowner survey (run with Owens Corning) found 74 percent of homeowners follow a recommendation from a neighbour, friend, or family member, and two-thirds gather three quotes. The referral earns you the call. It does not earn you the job. The job goes to whoever runs the better race against the other two quotes, and only 14 percent of contractors in that survey said they respond to a new enquiry in under five minutes.

One famous number is missing from this page on purpose. You will see the claim that 78 percent of buyers purchase from the company that responds first quoted across contractor marketing, usually credited to a survey nobody can produce. We could not trace it to any published study, so this playbook does not use it. The verified numbers above make the case without it.

Where do the minutes actually leak?

A roofing company's leads arrive through five doors, and each door leaks in a different place. Walk yours before you buy anything.

• Google Business Profile calls. Usually the highest-intent door, and usually routed to the office line or the owner's mobile. The leak: the owner is on a roof, the phone is in the truck, and the call ends in a voicemail box nobody checks until lunch. Invoca's platform data says 27 percent of calls to home services businesses are never answered at all.
• Website forms. The form submits to an inbox. The leak: an inbox is not an alert. Nobody's job is to watch it, evening submissions sit until morning, and there is no timestamp anyone trusts, so the delay is invisible as well as long.
• Lead marketplaces. Shared leads are sold to several companies at once, which compresses the race to minutes by design. The leak: a shared lead answered in two minutes can still beat an exclusive lead answered in two hours, so the marketplace gets blamed for a race your routing lost.
• Referral calls and texts. The best leads you get, arriving through the least formal channel. The leak: no logging, no owner, no record. A homeowner a neighbour vouched for expects a fast, personal response, and a text to the owner's personal phone during a pour is neither.
• Social and messaging apps. Enquiries land in a page inbox with notifications off, checked when someone remembers. The leak is pure neglect, and it concentrates on evenings and weekends.

Notice the pattern. Almost none of these leaks are effort. Each door has a different owner or no owner, the highest-value leads arrive when the crew is busiest or the office is closed, and callers pushed to voicemail almost never leave a message: Invoca's data puts it under three percent. The fix is plumbing and rules, not motivation.

Stage one: one intake, one timestamp

Every door gets wired into a single intake stream, so a Profile call, a form fill, a marketplace lead, and a referral text all land in the same place, in the same format, carrying the same fields:

• Who: name and the phone number to call back, mobile confirmed.
• What: leak, storm damage, replacement, repair, in the homeowner's words.
• Where: the address, because the callback and the inspection both need it.
• Which door: the source, recorded automatically, because the scoreboard later in this playbook judges each source separately.
• When: a system-generated timestamp. This is the timestamp the five-minute standard is measured against, and it is the only one nobody can edit after the fact.

The plumbing, in build order:

1. Point every form at the system, not an inbox. Your form tool writes the lead into your CRM and fires an alert in the same second. Whatever CRM you already run is fine. What matters is that the lead exists as a record with a timestamp before any human touches it.
2. Give each call source its own tracked number. Call tracking numbers (the CallRail pattern: one number per source, swapped on the website, calls recorded) keep the source attached to the call, so "the Google calls convert and the marketplace calls don't" becomes a measurement instead of an argument.
3. Parse marketplace leads into the same stream. Marketplace email notifications get parsed or integrated into the CRM automatically. A lead that lives in a marketplace app, checked twice a day, has already lost its race.
4. Send the holding text in seconds. An automated text within the first minute, in your company's name, that says a person will call within minutes, and means it. Its only job is to stop the homeowner calling the next company while your routing does its work. It does not ask questions, it does not qualify, and it never claims to be a person.
5. Flag the malformed, never drop them. A lead with a mistyped digit or a missing field gets flagged for a human to rescue. A lead the system silently discards is worse than no system at all.
6. Test it like a homeowner. Submit a test form at 7 p.m. from your own phone. Time what happens. This test is part of the build, and it repeats weekly.

Stage two: the callback, inside five minutes

Intake buys the minutes. Routing spends them. Two mechanisms do the work: a duty roster and an escalation rule.

The duty roster. One named person is first call for each day, with a named backup. Not "the office," not "whoever's free": a person, published a week ahead, covering the hours leads actually arrive, including evenings. When the roster says your name, your phone is the company's phone.

The escalation rule. The first phone rings for ninety seconds. No answer, the call rolls to the backup automatically. Still no answer, it rolls to the after-hours coverage in stage three. Ninety seconds is the threshold that makes the five-minute standard survivable: a roster without an escalation rule is a hope, and hopes do not answer on storm nights.

The call itself follows a five-beat structure. It is a booking call, not an estimate call, and it should take four minutes:

1. Open with their words. "You wrote that water is coming through the kitchen ceiling." This proves a human read the enquiry, and it is the fastest trust a cold callback can build.
2. Triage urgency. Active leak or interior storm damage first, then replacement, then repair. Urgency decides what happens next, not the order leads arrived in.
3. Ask two or three questions, no more. Address and access, whether an insurance claim is open, the timeline they are hoping for. Everything else belongs to the inspection.
4. Book the inspection on this call. Offer two concrete times. "We'll call to schedule" is where booked jobs go to die.
5. Confirm before you hang up. A text, sent while you are still talking, with the time, the address, and the name of who is coming. The homeowner leaves the call with an appointment and a person attached to it.

The caller works from the intake record, not from memory: source, the homeowner's own description, the full thread of any automated texts. If the duty person is calling from a truck between jobs, that record is their brief.

If nobody answers, the cadence is part of the system, not an afterthought. Velocify's 2013 contact-strategy study of roughly 3.5 million leads found 93 percent of leads that ever convert are reached by the sixth call attempt, and half of all leads never receive a second call at all. Spread six attempts across the first day and the two after it, mixing calls and texts, then stop. Speed wins the first touch; structured persistence wins the ones who were in the shower.

Stage three: after-hours coverage that is actually coverage

Homeowner expectations are more forgiving than the folklore suggests: in Roofing Contractor's 2026 survey, 39 percent of homeowners expected contact the same day and most of the rest within one to two days. You are not racing their expectations. You are racing the other two quotes, and storms do not keep office hours. Four coverage options, cheapest first:

1. The holding text plus a first-morning callback. Honest about what it is: not coverage, a promise. The overnight lead gets the automated text in seconds and the duty person's first call of the morning, in timestamp order. This is the floor, and it already beats a voicemail box.
2. An in-house on-call rotation. The roster extends into the evening with the same escalation rule, one person per night, compensated for carrying the phone. Works while after-hours volume is a handful of leads a week and someone will actually carry it.
3. A live answering service. Answers in your name, runs beats one and two of the script, books the inspection into your calendar, and warm-transfers anything with emergency language to the on-call human. The decision point is arithmetic: when after-hours leads per week times the jobs they represent outweighs the per-minute bill, the service is cheaper than the rotation's stipends and resentment.
4. An AI receptionist. Answers instantly, collects the intake fields, books into the same calendar. It earns its place only under the handoff rules in the next section, and only if its transcript lands in the same intake record the human would have written. An AI that takes messages into a separate silo is a voicemail box with better diction.

Whichever option you run, storm mode is a separate, pre-agreed state: when a named storm hits your service area, the full roster goes on, the script shortens to triage and book, and follow-ups queue by timestamp for the morning. Decide this in the calm. Nobody designs a surge protocol well during the surge.

The handoff rules: where automation stops

Automation in this system does exactly three things: it acknowledges, it routes, and it books. The moment any of these lines is crossed, a human owns the lead and the automation goes quiet:

• The homeowner replies. Any reply to any automated touch ends the automation. A person continues the thread, with the transcript in front of them. The homeowner never explains themselves twice; that is the failure this rule exists to prevent.
• Emergency language appears. Active leak, water inside, storm damage, anything unsafe: immediate human call, not a booked slot tomorrow.
• The enquiry is commercial or multi-property. Different conversation, different stakes. Human from the first touch.
• Two touches, no reply. Automation stops and the persistence cadence from stage two takes over. A third automated message is not persistence; it is noise with your name on it.
• The system is unsure. A malformed address, a number that looks like a landline, a form filled with half a story: flagged to a human, never guessed at.

And the rule that holds the rest together: a lead never receives two first touches. One record, one thread, one owner at a time. When the human takes over, the record shows everything the automation did and said, so the homeowner meets one company, not a relay race.

Where this system breaks

Build the guards with the system, because these failures are the normal ones:

• The routing app is logged out. The duty phone updated, the password changed, the alerts died silently three weeks ago. Guard: the weekly test call on the scoreboard below. If the test fails, the system is down, and it is treated that way.
• A source's race is unwinnable. Some shared-lead sources reach homeowners whose first responder is structurally faster than you. Guard: the scoreboard separates speed and booking rate by source. Two months of losing a source's race at full speed is evidence about the source, and it ends arguments that anecdotes cannot.
• The automation texts a landline. Guard: the form asks for a mobile number explicitly, and a flagged number falls back to a call-only cadence instead of three texts into the void.
• The booked work outruns the crews. This system books inspections, not start dates, and the script never promises a start date. If inspection capacity is the real constraint, that is a capacity problem, and it has a different fix than this playbook.
• Nobody believes the timestamps. Guard: the intake timestamp is system-generated and uneditable. The standard is measured against it or the standard is theatre.

What does it cost to run?

Three cost lines, all of them smaller than the leak they close:

• The tools. A CRM you likely already pay for, call tracking priced per number and per minute, the automation that moves leads between them, and whichever coverage option you chose in stage three, priced per minute or per month. Price them in your market; the categories are the budget, and any vendor who will not itemize them is telling you something.
• The owner's time. Fifteen minutes a week for the review below, plus roster management. That is the real cost, and it is why the review is fifteen minutes and not an hour: a ritual that survives busy season beats a thorough one that gets skipped.
• The comparison to run before you spend anything. Your monthly lead spend, against one month of this system's running cost, against the average job it protects: $17,631 for a residential replacement (Verisk, 2026). One rescued job a month is the bar the system has to clear, and it is a low bar at roofing job values.

If you are pricing the build itself rather than the running, the four ways to buy a system like this one, and what each really costs in year one, are in how much AI implementation costs for a small business.

Where doing it yourself stops making sense: when the roster becomes the job, when a second crew or a second service area makes the routing rules branch faster than you can maintain them, or when the weekly review keeps getting skipped because nobody owns it. The system is not complicated. Keeping it true, every week, in storm season, is the work.

The scoreboard: what to watch and the weekly review

Measure five numbers, weekly, split by source. Use medians, not averages: InsideSales' 2014 audit found a median first response of three hours and eight minutes against an average of 61 hours, because a tail of days-late responses poisons every average it touches.

1. Median time to first human contact. The standard, against the intake timestamp.
2. Share of leads touched by a human within five minutes. The headline number. Watch it by hour band; the evenings will show you your real coverage.
3. Live-answer rate on inbound calls.
4. Contact rate and booked-inspection rate, so speed never gets credited for conversations that go nowhere.
5. Booked rate inside versus outside five minutes, on your own leads. This is the number that replaces every borrowed benchmark in this playbook, and it takes one quarter to build.

The weekly review is fifteen minutes, same time every week, and it never agenda-creeps:

1. Pull the five numbers.
2. Find the single biggest leak: the slowest source, the dead hour band, the door with no owner.
3. Fix exactly that one, this week: a routing rule, an alert, a roster gap.
4. Place one test call and one test form, and time them.
5. Write the numbers down where last week's are. The trend is the management tool; any single week is weather.

Monthly, re-run the audit the honest way: pull twenty leads, read their timestamps, compare what the system says against what the CRM shows. The day your own inside-five-minutes booking rate is measured, the research sections of this playbook become background reading. That is the point of them.

For the rest of the system this plugs into, the growth systems for roofing companies hub covers being findable in the first place, the electrician quote follow-up playbook runs the same discipline on sent quotes in a sister trade, the venue enquiry response time piece tracks the same race in events, and ranking is not enough anymore makes the case that none of this matters if the phone never rings. One step earlier still, our audit of what ChatGPT recommends measures whether an engine names your company at all when a homeowner asks it for a roofer.

Key Takeaways

• Time to first human contact is the variable most roofing companies have never measured, and the verified research puts the cliff inside the first five minutes, not the first day.
• The standard is a briefed human inside five minutes. Automation holds the homeowner and arms the human; it never impersonates one.
• Every lead source leaks in a different place, and each leak is plumbing: one intake, one uneditable timestamp, one owner per lead.
• The callback is a four-minute booking call with a five-beat structure, and persistence is a designed cadence of up to six attempts, not a second try if someone remembers.
• Five numbers, fifteen minutes, one fix a week. The weekly review is what makes the standard survive storm season.

FAQ

How fast is fast enough for a roofing lead?

Five minutes to a human is the standard this playbook sets, and the research supports it as a direction rather than a magic threshold: the decay curve starts immediately and is steepest early. XANT's 2021 data found conversion eight times higher with a first call inside five minutes than after six or more; the 2007 study found the odds falling from the first five minutes onward. Treat five minutes as the line you staff and route against, and measure your own inside-versus-outside booking rate to see what it is worth in your market.

What is a realistic conversion rate on paid roofing leads?

We have not found a method-disclosed benchmark worth repeating, and the vendor percentages in circulation are sales material, not data. What is measurable is your own: the same source, the same month, split by whether a human touched the lead inside five minutes. Run that split for a quarter before you renegotiate with any lead seller; it changes what the numbers mean.

Should I answer calls myself or use an answering service?

Either beats voicemail by an enormous margin, since under three percent of callers sent to voicemail leave a message at all (Invoca). The choice is volume: an owner or a small rotation can carry a handful of after-hours leads a week; past that, a live service running your script's first two beats and booking into your calendar is cheaper than the jobs the rotation starts dropping.

Do shared or resold leads convert at all?

Yes, when you win the race they are designed to create. A shared lead is the same enquiry in several companies' intakes at the same second, so speed matters more with shared leads, not less. If your scoreboard shows one source's race being lost at full speed for two straight months, that is a sourcing decision to make with data, not a reason to answer everything slower.

How do I handle leads that come in overnight?

The holding text fires in seconds, the emergency language rule escalates anything urgent to the on-call human immediately, and everything else gets the duty person's first call in the morning, in timestamp order. What overnight leads cannot survive is an inbox and a voicemail box: those are the two places the 42-hour average lives.

Why isn't the 78 percent first-responder statistic in this playbook?

Because nobody can produce the study behind it. It is attributed to a survey that has no published report or methodology anyone has been able to locate, and it has been copied from blog to blog until the copying looked like sourcing. The figures this playbook does cite are traced to named studies with stated samples, in Sources below. If the 78 percent study ever surfaces, we will cite it with its method attached.

Sources

• The Short Life of Online Sales Leads, Harvard Business Review, March 2011: the 2,241-company audit, the 42-hour average, and the within-an-hour qualification advantage.
• Lead Response Management study (Oldroyd for InsideSales, 2007): the 100x contact and 21x qualification odds between five and thirty minutes, from 15,000+ leads and 100,000+ call attempts. Traced to the original study via the source-by-source compilations at expertise.ai and ainora.lt, both checked October 2026.
• Lead Response Management study (XANT, formerly InsideSales, 2021): 5.7 million inbound leads, 2018 to 2020 data; the eight-times conversion finding, the sub-one-percent five-minute attempt rate, and the 57 percent week-plus first attempts. Traced as above.
• How Much Missed Sales Calls Cost Home Services Businesses, Invoca: 27 percent of calls to home services businesses unanswered; under three percent of voicemail callers leave a message.
• Lead Response Survey (Drift, 2017) and the 1,000-company demo-form test (RevenueHero, 2024): the seven percent five-minute response rate and the 635 non-replies. Traced as above.
• The Ultimate Contact Strategy (Velocify, 2013): the 391 percent first-minute conversion lift later popularized by Chili Piper's guides, the six-attempt reach finding, and the half of leads never called twice. Traced as above.
• Roofing Reality Check: Risk Is Rising Even in Quiet Storm Years, Verisk 2026 U.S. Roof Report: the $17,631 average residential replacement cost for 2025.
• The Homeowner's Roofing Journey in 2026, Roofing Contractor: the homeowner survey (with Owens Corning) behind the 74 percent recommendation figure, the three-quote behaviour, and the 14 percent of contractors responding in under five minutes.

We source every figure here the way our own editorial policy describes: traceable to a named source with its method attached, or left out. Read more about who writes and fact-checks this blog.