How Roofers Scale Past Referrals: The 4 Channels, Ranked

Referrals are the best leads you will ever get and they cannot scale — volume is capped by your install base times a referral rate you don't control. Four owned channels grow past that ceiling. Ranked by cost per closed job: direct outreach to storm-hit owners, storm response inside the claim window, canvassing aimed by property data, then inbound — which costs the most up front and the least forever after. Shared marketplace leads finish last at roughly $833 per closed job on generous assumptions, and never improve.

The referral ceiling is math, not effort

Every referral job is downstream of a past job. If you've installed 300 roofs and a great year refers at 8-10%, that's 24-30 referred jobs — and no amount of hustle moves that number quickly, because the input is your history, not your activity. Referrals also cluster geographically and demographically: the same neighborhoods, the same price band, the same storm cohort aging out together. A referral-only roofing company isn't a growth company; it's an annuity on its own past work. Good annuity — keep it — but the ceiling is structural.

Why bought leads don't fix it

The reflex move is buying leads, and the math usually disappoints: marketplace leads run $50-200 each and are sold to as many as four contractors at once, so you're paying for a phone race you win a quarter of the time. Close rates on shared leads sit far below referral close rates, which means the real cost per JOB — not per lead — often lands north of $800 before you've priced the rep hours burned on the losers. Bought leads are a bridge, not a channel: you rent demand and own nothing afterward.

The four channels, ranked by cost per closed job

Growth past referrals comes from channels you own. Ranked below by what each one costs to produce a signed contract — not a lead, not an appointment, a job. Referrals and shared leads are included as reference points at the top and bottom, because the whole argument lives in the gap between them. Dollar figures are a worked model on the assumptions in the next section; the ordering is what holds, not the decimals.

• Direct owner outreach ranks first on cost because it is mostly your time, not your money — calling or texting the verified owner of a storm-hit, aging-roof property turns prospecting into triage. Owner-occupancy filtering alone removes the roughly one-third of doors that legally cannot say yes to you, and a phone append costs pennies per record against $50-200 for a shared lead.
• Storm response ranks second and produces fastest, but you do not control the schedule. After confirmed hail, every roof inside the footprint has a real, insurable need and a filing deadline. The contractor who reaches those doors inside the claim window with the storm data already in hand wins jobs referrals never touch — and the ones who show up in week three are selling against a signed contract.
• Intelligent canvassing costs more per job than the first two and is the only one of the four that scales linearly with headcount — which makes it the growth engine. Door-to-door works when the doors are chosen: homes ranked by roof age, storm exposure, and owner status convert on signal instead of volume, and the rep's day is routed instead of wandered.
• Inbound is last on speed and first on durability. Reviews, local search presence, and being findable when a neighbor asks the internet instead of a friend. It is the only channel on the list whose cost per job falls every year, and the only one that still produces when you stop working it.

Run the arithmetic on your own shop

The ranking above is worked on one set of assumptions. Yours will differ, and the point is that you can check it in ten minutes rather than trust a table on the internet. The formula for every channel is the same: cost per closed job = (what you spend to produce the opportunities) divided by (opportunities x your close rate on that channel). Two things go wrong when roofers skip this. First, they measure cost per LEAD, which flatters every paid channel — a $125 lead looks cheaper than a rep's afternoon right up until you divide by a 15% close rate. Second, they use one blended close rate across all channels, which is the single most expensive mistake in the list, because a referral closes near 50% and a shared lead closes nowhere near that.

• Shared leads, as modeled above: $125 average lead cost divided by a 15% close rate = $833 per job. Fifteen percent is generous for a lead sold to four contractors at once — at 10% it is $1,250. Run yours; the number is usually worse than the one you carry in your head.
• Direct owner outreach, as modeled: 200 phone appends at roughly $0.10 each is $20, plus six rep-hours at a $25 loaded rate is $150. Call it $170 producing two jobs, and you are at $85 per job. The lever is not the append cost, it is whether the 200 records were chosen well.
• Storm response, as modeled: one rep working 60 hours inside a claim window at a $25 loaded rate is $1,500, producing 8-12 jobs in a genuinely hit neighborhood. That is $125-190 per job, plus whatever your storm data costs amortized across the event.
• Canvassing, as modeled: the same $25 loaded rate against blind knocking at roughly one sale per 40-plus doors, versus ranked knocking where the denominator falls because the reps stop spending afternoons on streets with new roofs and absentee owners. The cost per job is entirely a function of that denominator, which is why targeting is the whole game.
• Track cost per job by channel monthly, in a spreadsheet if that is what you have. The ranking will shift with your market and your reps — what will not shift is that the channels where you own the asset beat the ones where you rent demand.

The common thread: intelligence before activity

Every channel on that list works in proportion to how well you choose targets before spending effort. Canvassing fails as random door-knocking and works as ranked door-knocking. Storm response fails as chasing every cloud and works as verified storm footprints with claim-window countdowns. Outreach fails as cold lists and works as owner-verified, storm-hit properties. This is the actual lesson of scaling past referrals: referrals came pre-qualified by a happy customer — every replacement channel needs its own qualification layer, and that layer is data. It's also, transparently, the layer Curbsight sells: every property in a territory ranked 0-100 by Curbsight's proprietary scoring, with the canvassing, routing, owner phones, and CRM attached so the intelligence turns into knocked doors and closed jobs in one system.

The first 30 days, in order

Sequenced so that each step funds the next one and nothing requires a budget approval to start. This is deliberately ordered cheapest-and-fastest first, which is the reverse of how most shops do it — the usual pattern is to buy leads in week one because it is the only step that requires no thought, and to still be buying them in year three.

• Week 1 - Measure what you already have. Pull your last 12 months of closed jobs and tag each one by how it originated: referral, canvass, storm, inbound, bought. Most shops discover their referral rate is not what they assumed and that one channel they barely fund is outperforming. You cannot rank channels you have never separated.
• Week 1 - Compute your close rate per channel, not blended. This is the input that makes every other number in this article real for your shop rather than illustrative.
• Week 2 - Build one list of 200 owner-occupied homes with roofs old enough to matter in a storm-exposed area you already work. Roof age from assessor records, owner-occupancy from the same, storm exposure from public NOAA history. Call or text them. This is the cheapest channel on the ranking and it needs no new hire and no software you do not already have access to.
• Week 3 - Set a storm trigger. Decide now what you do in the 72 hours after confirmed hail in your service area: who stops what they are doing, which neighborhoods get worked first, what the door script is. Contractors lose these jobs to the shop that had already decided, not to the shop with better reps.
• Week 4 - Fix the canvassing denominator before adding reps. Adding a rep to a blind map multiplies the waste. Rank the doors first - by roof age, by storm exposure, by owner status - and the same rep count produces more contracts without working longer.
• Ongoing - Start the inbound clock now even though it pays in month nine. Ask every closed customer for a review the day the job passes final inspection, while the roof is new and they are happiest. It is the highest-conversion moment you will ever have with them and most shops ask three weeks late, by email, and get nothing.

What a realistic ramp looks like

A 3-5 rep residential company that adds intelligent canvassing plus storm response typically sees the mix shift inside one season: referrals keep producing at their fixed rate while canvassed and storm work grows with rep count — the thing referrals can never do. The discipline that matters most is measuring cost per JOB by channel, not cost per lead: referrals will stay the cheapest, owned canvassing lands next, storm response spikes cheapest-per-job in the weeks after an event, and shared leads almost always finish last. Scale the channels where you own the asset — the territory, the data, the review base — because those compound while rented demand resets to zero every month.

Frequently asked questions

Sol Zendejas-Smith — Sol Zendejas-Smith worked home-services sales and roofing in Oklahoma before founding Curbsight in 2026. He built the platform he wished he'd had at the door.

• The real cost of a roofing software stack →
• Best roofing CRM software in 2026 →
• How many jobs before paper breaks →
• Roofing use case →

Ready to see your scored territory?