How to Predict Which Homes Need a Roof Replacement (A Field Playbook)

Every roofing company runs on the same hidden math, whether the owner has ever written it down or not. You knock a street, you mail a ZIP, you buy a list, and somewhere in that pile of doors is a small number of houses that are genuinely worn out and a much larger number that are nowhere close. The crews that grow are the ones that get good at telling those two groups apart before they spend the gas, the postage, and the payroll. The crews that stay flat treat the whole street as one undifferentiated blob and let the law of averages decide their week.

Predicting which homes need a roof replacement is not a parlor trick and it is not magic. It is a discipline built out of a handful of signals that any sharp roofer already half-knows: how old the roof probably is, what weather it has actually absorbed, what the surface looks like from the curb, and what the property records quietly tell you. The trick is to stop using those signals one at a time, by gut, and start stacking them into a repeatable read you can hand to a green canvasser and trust.

Below is the working method. It is the same logic a good estimator runs in his head as he drives a neighborhood, broken into steps you can train, measure, and scale. We will cover the age signals that matter, the storm-wear signals most people botch, the ground-level curb read, how property and imagery data fill the gaps, how to combine everything into a simple score, and how to turn that score into a route your crew can actually run on a Tuesday morning. We will also be honest about where prediction ends and the ladder begins, because anyone who tells you they can grade a roof from a satellite with certainty is selling you something.

What "due for replacement" actually means

Before you can predict it, define it. A roof is due when the cost of keeping it alive starts losing to the cost of replacing it, or when a single event pushes a marginal roof over the edge. In practice, three different situations send a homeowner to you, and they need different reads:

• Age-out. The shingles have simply run out of service life. An asphalt three-tab from the early 2000s, a builder-grade architectural shingle from a 2006 subdivision, a roof that has baked through twenty southern summers. No single dramatic failure, just a surface that has lost its granules, its flexibility, and its sealant grab. These are the bread-and-butter jobs that never make the news.
• Storm wear. A roof with years of life left gets shortened or ended by hail bruising, wind uplift, or debris impact. The homeowner often has no idea anything happened. The damage is real but invisible from the driveway, and it frequently rides alongside an insurance claim.
• Defect or installation failure. Premature curling, blistering, a bad batch of shingles, improper nailing, or a chronic leak from flashing and penetrations. Younger than its age would suggest, failing for reasons unrelated to weather.

Most prediction effort goes into the first two, because they are the ones you can read at scale without being on the roof. The third usually surfaces through inspection or a homeowner complaint, not from the street. Keep the three buckets separate in your head, because a hail neighborhood and an age-out neighborhood get worked with completely different messaging.

The age signal: your single strongest predictor

If you could know only one thing about a roof before knocking, you would want its real installed age. Age is the backbone of every honest prediction, because shingle failure is overwhelmingly a function of time plus exposure. Asphalt shingles are an organic-and-mineral product that degrades on a fairly predictable curve once they pass the midpoint of their rated life.

Here is the part most reps get wrong: the year a house was built is not the age of its roof. A 1998 house may be on its second or third roof. A 2015 house may already have a hail-replaced roof that is newer than the structure. Tax records, Zillow, and county GIS all show year built, and that number silently hides every re-roof that ever happened. Re-roofs are invisible to the public record, and re-roofs are exactly the events that reset the clock you care about.

So the goal is not to find the build year. The goal is to estimate the age of the current roof surface as a range. A range, not a date. Anyone who hands you "this roof was installed on March 12, 2009" from a desk is guessing with false precision. A defensible answer looks like "this roof reads 17 to 22 years old," and that range is enough to make a routing decision.

Rated life versus real life

Know the rated life of the common products so you can map an age estimate to a probability of failure:

The rated number on the wrapper is a warranty marketing figure measured under ideal conditions. Real service life in Texas, Oklahoma, Colorado's Front Range, or the Gulf Coast runs well short of it because UV, thermal cycling, and storm exposure compound. A 30-year architectural shingle in Phoenix is often visibly cooked by year 18. Use the realistic column, not the wrapper, when you build your model.

Adjust the clock for your climate

The same shingle ages at very different speeds depending on where it sits, and a prediction model that ignores climate will mis-call roofs by years. Build a simple regional multiplier into your age read so the same product gets graded "due" sooner in a punishing climate than in a mild one.

• High UV and heat (Arizona, Nevada, inland California, Texas). Thermal cycling and ultraviolet exposure break down the asphalt binder fast. Knock several years off the realistic service life. A roof here reads old early, and the granule loss tells you so.
• Hail and severe-storm corridors (the Plains, Colorado Front Range, North Texas, Oklahoma, the upper Midwest). Service life is driven less by slow aging and more by event wear. A ten-year-old roof here can be functionally finished if it caught two real hail seasons. Weight the storm component heavily.
• Hot and humid coastal (Gulf Coast, Southeast). Heat plus moisture plus wind. Algae streaking shows early, sealant strips fail, and high-wind events lift tabs. Watch the dark streaking and the leeward edges.
• Cold and freeze-thaw (Northeast, Mountain West, upper Midwest winters). Ice damming, freeze-thaw cracking, and snow load. Failures cluster at eaves, valleys, and the north slopes that never dry out. Roofs may look better longer here but fail at the details first.
• Mild marine (Pacific Northwest, coastal mid-Atlantic). Slower aging, more moss and organic growth, often the longest realistic service lives. Don't over-call age here; the curb read matters more than the calendar.

The practical move is a single multiplier per market. If your baseline architectural shingle reads "due at 18 years," a high-UV Texas market might use 15, a mild Oregon market 22. Calibrate the number against the real ages of roofs you have already replaced, and you will stop both over-pitching young roofs in mild markets and under-targeting cooked roofs in harsh ones.

How to estimate roof age without a date

You triangulate. No single source is reliable alone, so you stack:

1. Build year as a ceiling, not an answer. The roof cannot be older than the house. In neighborhoods built and never re-roofed, build year is a decent floor estimate. In storm-hit metros, treat it with suspicion.
2. Permit history. Many jurisdictions require a roofing permit and publish them. A 2011 re-roof permit on a 1994 house tells you the real story. Permit data is patchy and many re-roofs were done without one, but where it exists it is gold.
3. Aerial imagery over time. Historical aerial photos show the roof changing color and texture across years. A roof that was dark and crisp in a 2014 image and is streaked and faded in the current image has aged in plain view. A roof that suddenly went from worn to bright new between two image dates was re-roofed in that window.
4. Subdivision cohort. Tract neighborhoods were roofed in a tight window. If you confirm the age of a few roofs on a street through permits or a closed job, the un-re-roofed neighbors are likely the same vintage. Cohort logic is how you scale one data point into a block.
5. Visible weathering grade. The curb read, covered below, calibrates everything. Granule loss and color fade are the eye's version of a clock.

None of these is precise. Together they collapse the uncertainty enough to put a confident range on a roof. That range is the foundation everything else stacks on.

The storm signal: where most prediction goes wrong

Age tells you which roofs are aging out. Storm history tells you which roofs got their life cut short — and this is where most companies, even good ones, make a predictable mistake. They pull a hail map.

A hail map shows you where it hailed. It draws a swath, you pick the ZIP codes the swath crossed, and you flood the whole area with knockers. The problem is that a hail swath is enormous and crude, and within it the actual roof damage is wildly uneven. Two houses on the same street can have completely different outcomes from the same storm depending on:

• Hailstone size and density at that exact point, which varies block to block and is not captured by a swath outline.
• The direction the storm traveled and the wind that drove it. Hail comes in at an angle. The roof planes facing into the storm take the hits; the leeward planes may be untouched. A north-facing slope and a south-facing slope on the same house can grade differently.
• Roof pitch and orientation, which change the angle of impact and therefore the energy delivered to the mat.
• Shingle age and brittleness. The same hailstone that bounces off a three-year-old roof fractures the mat on a sixteen-year-old one. Storm and age are not separate signals; they multiply.
• Tree cover and surrounding structures that shield some roofs and funnel debris onto others.

This is why a ZIP-level hail map produces so much wasted knocking. You knock a hundred doors inside the swath and find that thirty roofs took real damage, forty are borderline, and thirty are fine, and you had no way to tell which was which before you walked up. The map told you the storm passed over the neighborhood. It did not tell you which roofs the storm actually wore out.

Modeling the storm on the roof, not over the ZIP

The better read treats each roof as a target the storm hit at a specific angle with specific energy. Instead of "this ZIP got hail," you want "this roof, at this pitch and orientation, sat under estimated 1.75-inch hail driven from the southwest, and it was already 15 years old." That is a per-roof read, and it ranks doors inside the swath instead of treating them all the same.

You can approximate this yourself with effort: pull the storm's path and reported hail sizes, note the prevailing wind direction during the event, and cross it against each roof's orientation and age. It is laborious by hand, which is exactly why most crews skip it and fall back to the lazy ZIP sweep. The payoff for doing it is that you walk up to the doors most likely to have real, documentable damage first, while the swarm of out-of-town crews is still knocking randomly.

A note on honesty here, because it matters legally and reputationally: a storm model gives you odds, not proof. Modeling tells you a roof was likely damaged enough to warrant a look. It is a reason to inspect, not evidence to wave at a homeowner or an adjuster. The proof is what your inspection documents on the roof itself. Treat the forecast as a targeting tool and let the ladder confirm it.

The curb read: grading a roof from the ground

This is the skill that separates a rep who can canvass from a rep who is just walking. You can learn an enormous amount about a roof from the street and the driveway without ever setting a ladder, and you can train it. Here is the ground-level read, roughly in the order your eye should run.

Color and granule loss

Asphalt shingles get their color and their UV protection from the mineral granules embedded in the surface. As a roof ages, those granules shed. The first tell is color fade — a roof that has gone from rich charcoal to a chalky gray, or from brown to a washed-out tan, has lost surface. The second is shine or sheen, where so many granules are gone that the black asphalt substrate shows through and the roof looks slick or oily in direct sun. A shiny asphalt roof is a roof near the end. Check the gutters and the splash zones at downspout outlets, too; a pile of granules that looks like coarse black sand is the roof telling you it is shedding fast.

Surface texture and shape

• Curling and cupping. Shingle edges lifting or the centers cupping up. This is moisture and age deforming the mat. Visible from the ground on the planes that face you, especially in raking afternoon light.
• Clawing. The opposite curl, where the middle lifts. Same story: an old, dried-out roof.
• Waviness or buckling lines. Long ripples across a plane suggest deck movement, moisture in the underlayment, or layered-over old roofs. Often a sign of a roof that was re-covered rather than torn off.
• Missing or lifted tabs. Gaps in the field, or tabs that have been wind-lifted and reseated crooked. Wind damage and age both produce this.
• Sagging ridge or deck dips. A roofline that is no longer straight points to structural or long-term moisture issues. That is a bigger conversation than a re-roof, but it is a strong "this house needs us" flag.

Flashing, penetrations, and the soft spots

The field of the roof is rarely where leaks start. Look at the valleys, the chimney and skylight flashing, the pipe boots, and the step flashing along walls. Cracked rubber pipe boots are one of the most common real failure points and are sometimes visible from the ground. Rust streaks running down from a flashing detail, dark staining around a chimney, or a tarp anybody threw up there are all reads you can make from the curb.

Calibrate your eye against known roofs

The curb read is only useful if it is consistent across reps. Build a calibration set: photograph ten roofs you have already gotten on and graded by hand, label each with its real condition and age, and use them to train new canvassers. Have a green rep grade those ten from the street, then show them the ground truth. Two afternoons of that and a new hire reads a roof far better than instinct alone. Without this calibration, "curb appeal grading" is just opinion, and ten reps will give you ten different answers.

What the ground read cannot tell you

Be disciplined about the limits. From the ground you cannot reliably assess hail bruising — those soft, granule-knocked dents that don't show until you are kneeling on the plane with a piece of chalk. You usually cannot see the back slopes. You cannot judge decking condition, underlayment, or attic ventilation, all of which affect the real recommendation. The curb read is a triage tool that tells you which roofs deserve a ladder. It is not a substitute for the inspection. Sell it to your reps that way, or they will overpromise on a doorstep and get embarrassed when they get up top.

The records read: property data that sharpens the picture

Before your crew burns a tank of gas, a surprising amount can be pre-filtered from a desk. None of it is conclusive, but it cheaply removes the obvious non-starters and flags the obvious targets.

The move is to use records to remove doors as much as to add them. A house that sold last year with listing photos showing a crisp new roof goes to the bottom of your list. A 1996 tract home with no re-roof permit on file and a storm in the area three years ago goes near the top. You are not trying to be certain. You are trying to spend your knocking on the highest-probability doors first.

Picking the right neighborhoods before you pick the right doors

House-by-house scoring is where the money is, but you still have to decide which neighborhoods are worth scoring at all. Reading the street is door-level work; reading the map is route-level work, and the companies that grow do both. A neighborhood that looks tempting on a drive-through can be a payroll sink, and a plain-looking subdivision can be a season's worth of jobs. Here is how to read an area before you commit a crew to it.

What a high-yield neighborhood looks like

• A tight build cohort in the replacement window. Tract subdivisions roofed in a two-to-three-year stretch are the dream, because when the cohort ages out, dozens of roofs come due together. A 2002-built subdivision in 2026 is a wall of architectural shingles all hitting end of life at once. One confirmed age reads across the whole tract.
• Owner-occupied, stable ownership. Long-tenure owners make the decision and live with the roof. Heavy rental or absentee ownership slows everything down because the decision-maker isn't behind the door.
• Middle-market homes. Not so cheap that a homeowner defers indefinitely, not so high-end that they have a roofer on retainer. The meat of the market is the middle.
• A real, documentable storm history if you are running the storm play. One significant hail or wind event in the last few years, not a vague "we get storms here."
• Low recent re-roof saturation. If half the neighborhood already shows fresh roofs in recent aerials, the cohort has already been worked. Someone got there first; move on.

What a low-yield neighborhood looks like

• New construction or recently built. Roofs are years from due. Beautiful, and a complete waste of your week.
• Already-worked storm areas where the fresh roofs in the aerials tell you the swarm already swept it.
• Mixed-vintage, custom-home areas with no cohort, where every roof is a different age and product. You can still find jobs, but you can't scale a route; it's one-off prospecting.
• High HOA-control or deed-restricted communities that route all roofing through an approved-vendor process. Know these before you knock them.

A quick neighborhood scoring pass

Before you score individual doors, rank candidate neighborhoods on four cheap reads: dominant build cohort age, re-roof saturation visible in current aerials, ownership stability, and storm history. A neighborhood that scores well on all four gets the full house-by-house treatment first. This two-tier approach — rank areas, then rank doors inside the best areas — is how you avoid scoring five hundred homes in a subdivision that was already picked clean.

Putting it together: a simple roof-replacement probability score

Now stack the signals into one number so you can rank a street instead of arguing about each house. You do not need a data scientist for the first version. A weighted 0–100 score that any sales manager can compute beats the gut-feel blob you are replacing. Here is a starter model you can adapt to your market.

A starter scoring model

Assign points, weighting age and storm wear most heavily because they are the strongest predictors:

Age component (up to 45 points)

• Estimated roof age within 3 years of rated end-of-life or beyond: 45
• Within the "starts reading due" window from the table above: 30
• Mid-life (roughly 8–13 yr on an architectural roof): 12
• Clearly young (under ~6 yr or recent re-roof confirmed): 0

Storm-wear component (up to 30 points)

• Roof orientation faced into a significant hail/wind event while already mid-to-late life: 30
• Inside a real storm swath but favorable orientation or younger roof: 15
• No notable storm history: 0

Curb-condition component (up to 20 points)

• Multiple end-of-life tells (sheen, curling, granule piles): 20
• One or two moderate tells (fade, isolated lifting): 10
• Looks sound from the street: 0

Records flags (up to 5 points, can go negative)

• Old build year, no re-roof permit on record: +5
• Confirmed recent re-roof (permit, listing photo, fresh aerial): −40 (this should sink the door regardless of other signals)

Sum it. A house at 75+ is a knock-first door. A house at 40–74 is a worth-a-look door for a second pass. A house under 40, and especially any confirmed recent re-roof, is a skip. The exact thresholds are yours to tune against your close data, and you should tune them — feed back which scored doors actually became jobs and adjust the weights every quarter.

A worked example

Take three houses on the same 1999-built cul-de-sac, after a hail event eighteen months ago that tracked through from the southwest with reported 1.5-inch stones.

• House A: 1999 build, no re-roof permit, architectural shingle, west and south planes face the storm track, curb read shows fade and a little granule in the gutter. Age 30 + storm 30 + curb 10 + records 5 = 75. Knock first.
• House B: 1999 build, but a 2019 re-roof permit on file and a bright, crisp roof in the latest aerial. Age 0 + storm 15 + curb 0 + records −40 = −25, floored to skip. Do not waste the door.
• House C: 1999 build, no permit, but the storm-facing planes are on the back and the visible front planes look mid-life with light wear. Age 30 + storm 15 + curb 10 + records 5 = 60. Second pass.

Same street, same storm, same build year, three completely different decisions. That is the entire point. The blob says "knock the cul-de-sac." The score says knock A now, swing back to C, and skip B entirely — which saves you a door, a pitch, and the small reputational hit of pitching a roof somebody just replaced.

Where RoofPredict fits

Everything above is doable by hand. It is also slow, and the by-hand version falls apart at scale — nobody is going to triangulate roof age, model storm orientation, and curb-grade five hundred houses with a spreadsheet before Tuesday. That gap between "the right way to read a street" and "the way anyone actually has time to do it" is the problem RoofPredict was built to close.

RoofPredict scores the roofs in an area the same way the playbook above does, automatically. It estimates roof age from aerial imagery as a range — not a fake exact date, a defensible range like 17 to 22 years — and it models storm physics per roof rather than per ZIP. Instead of handing you a hail swath and wishing you luck, it accounts for each roof's orientation and the storm's actual track and energy, then pairs that with the age read to rank the doors on a street by how likely they are to be genuinely worn out. You get an ordered list: knock these, look at these, skip these.

What that changes day to day: your canvassers start at the top of a ranked street instead of the corner house. Your mail goes to the roofs that read old, not the whole carrier route, so you stop paying postage on new construction. A green hire who could not yet grade a roof from the curb still knocks the right doors, because the ranking is doing the triage the calibration set teaches a veteran to do. And your CRM full of old estimates can be re-scored, surfacing the homeowners whose roofs have quietly aged into the replacement window since you last quoted them.

Now the honest limits, because a tool that oversells itself gets a roofer burned on a doorstep. RoofPredict gives you odds, not proof. The age is a range, not a certificate. The storm model tells you a roof was likely worn or hit, which is a reason to get up there and document — it is not evidence of damage and it is not a claim. It does not measure the roof for a takeoff, it does not identify the exact shingle product, and it does not decide anything about insurance: the roofer documents the conditions and writes the estimate, the insurer decides coverage, and the homeowner owns their claim. What it does is the one thing that eats your week if you do it by hand — point the crew at the right houses first. It is targeting and routing, not a substitute for the inspection or the relationship. Used that way, it makes the playbook above runnable at the size of your actual territory.

From prediction to a route your crew can run

A score is useless if it dies in a spreadsheet. The last mile is turning the ranking into a day of work. Here is the operational handoff.

Build the route around the doors, not the geography

The instinct is to knock a neighborhood corner to corner because it is efficient walking. The better instinct is to knock the high-score doors in a tight cluster and let the low scores fall out. If a street has six 75+ doors, those six are the route, and the new roofs in between are walked past on purpose. You will cover less ground and book more inspections, because every conversation is with a homeowner whose roof is actually due.

Sequence the day

1. Pull the ranked list for the target area the night before. Reps should hit the field knowing their first ten doors, not deciding on the sidewalk.
2. Lead with the highest-probability cluster while energy is fresh and the day is young — these are your best shots at a booked inspection.
3. Work second-pass doors (the 40–74 band) in the gaps and on the back end of the day.
4. Log every door's outcome back against its score. No-answers, not-interested, booked, and the eventual close. This is the data that retrains your weights.

Match the message to the bucket

An age-out neighborhood and a storm neighborhood need different opening lines, and a green rep who uses the storm pitch on an age-out street sounds like every other knocker.

• Age-out door: "I've been working roofs on this street and yours is reading like it's getting up there in age. I'd rather you hear it from someone honest now than find out with a leak in January. Mind if I take a quick look?" No urgency theater, no storm story that isn't true.
• Storm door: "You had a real hail event come through about a year and a half ago, and roofs facing the way yours does tend to take the brunt of it. A lot of that damage doesn't show from the driveway. I'd like to get up there and document what's actually going on, no charge to look." Then document, hand the homeowner the findings, and stay completely out of telling them what insurance will or won't do — that is the adjuster's call and the homeowner's claim.

Close the loop with documentation

Whatever the bucket, the inspection is where prediction becomes a real recommendation. Photograph everything, grade honestly, and give the homeowner a clear written read of the roof's condition and remaining life. If there is storm damage, document the conditions and the estimate and let the homeowner decide whether to file. You are the one who got on the roof and wrote down what was there. Keep your role to exactly that, and you build the kind of reputation that makes the next street easier to work.

The economics: why prediction pays for itself

Targeting is not a nice-to-have; it changes the unit economics of your whole outbound. Work through the math once and it becomes obvious why the blob approach quietly bleeds money.

Say a canvasser works 50 doors a shift and gets a conversation at maybe one in four. On a random street where one roof in five is genuinely due, only a fraction of those conversations are even with the right homeowner, and the rest are polite no-thank-yous from people whose roofs are fine. Now flip it: if your ranked list puts the due roofs first, the same 50 doors are concentrated on homeowners who actually need you. You have not added a single hour of labor. You have changed who gets the knock, and the booked-inspection rate climbs because the conversations are relevant.

The same logic governs mail. Direct mail to a full carrier route sends a postcard to every new build, every recently re-roofed house, and every roof with a decade of life left. If a meaningful slice of a route is roofs that don't need you, every one of those pieces is postage and print spent on a guaranteed non-buyer. Mailing only the homes that read old shrinks the drop, raises the response rate per piece, and lets you mail the right neighborhoods more often for the same budget.

A simple way to track whether your targeting works

You cannot improve what you don't measure. Track these per neighborhood and per campaign:

Run a targeted neighborhood and an untargeted one side by side for a month and compare cost per closed job. The gap is the dollar value of prediction, and it is usually large enough that the only question left is why you waited. The companies that win treat this scoreboard as seriously as they treat their production schedule.

Don't forget the safety and reputation math

There is a cost the spreadsheet misses. Every roof your crew climbs to inspect carries fall risk, and roofing is consistently among the most dangerous trades — follow OSHA fall-protection practices on every inspection, no exceptions. Fewer wasted climbs on roofs that were never going to be jobs means less exposure for the same revenue. And every homeowner you pitch a brand-new roof costs you a little credibility on the street; getting the read right protects the reputation that makes the next neighborhood easier. Prediction is partly a labor-efficiency play and partly a trust-preservation play, and both show up in the numbers eventually.

The mistakes that quietly cost you the most

A few patterns separate the crews that grow from the ones that grind. Most of them are about discipline, not talent.

• Trusting build year as roof age. The single most common and most expensive error. You pitch new roofs, you skip re-roofed targets, and you wonder why your numbers are random. Build year is a ceiling, never an answer.
• Sweeping the whole ZIP after a storm. The hail map feels like data, but it is the bluntest possible instrument. You burn payroll knocking roofs the storm never really touched, and you arrive at the genuinely damaged doors exhausted and behind the out-of-town crews.
• Treating every door in a swath the same. Orientation, pitch, and age decide who got hurt inside the storm. Ignoring that is leaving your best doors un-prioritized.
• Letting curb grading be opinion. Without a calibration set, ten reps grade ten ways and your "data" is noise. Train the eye against known roofs or don't claim to grade from the ground.
• Overpromising on the doorstep. A rep who tells a homeowner "you have hail damage" before getting on the roof is writing a check the inspection has to cash. Predict likely, then go prove it. Forecast is odds, never proof.
• Never feeding outcomes back. If you never compare which scored doors actually became jobs, your model never improves and your targeting stays a guess with extra steps.
• Confusing targeting with claims handling. Your job is to find the worn roofs and document conditions. The insurer decides coverage and the homeowner owns the claim. Crews that blur that line invite trouble they don't need.

A 30-day plan to make this your default operating system

If you want to move from reading this to running it, here is a concrete on-ramp.

Week 1 — Build the calibration set and define your buckets. Photograph and hand-grade ten to fifteen roofs you have already gotten on. Write down each roof's real age and condition. Define your age-out window for the shingle types common in your market using the realistic service-life figures, not the wrapper warranties.

Week 2 — Assemble your data inputs. Get access to your county's assessor and permit data. Pull the last few years of significant hail and wind events in your service area from public storm records. Identify the tract neighborhoods and their build cohorts. Decide whether you are doing the storm-orientation read by hand or bringing in a tool that scores roofs per-house so you are not triangulating five hundred homes by yourself.

Week 3 — Score a real neighborhood and route it. Pick one target area. Score every house with the starter model. Build a ranked route around the high-probability clusters. Send your reps in with the list and have them log every outcome against the score.

Week 4 — Read the feedback and tune. Look at which scored doors booked inspections and which became jobs. Were your 75+ doors actually due? Did anything you scored low surprise you? Adjust the weights. Update the canvasser calibration set with any roof you got on that read differently than expected. Then run it again on the next neighborhood.

Do that for a quarter and the read stops being a special project and becomes the way your company decides where to send people every single morning. The math that was hiding in every roofer's gut becomes a system you can hand to a twenty-two-year-old on his first week and trust to point him at the right doors.

The bottom line

Predicting which homes need a roof replacement comes down to stacking four reads — real roof age as a range, storm wear modeled on the actual roof rather than smeared across a ZIP, an honest curb grade, and what the property records quietly confirm — into one score that ranks a street. The roofs that are aging out and the roofs a storm genuinely wore down rise to the top; the new ones and the recently re-roofed ones drop out. You knock the right doors, you mail the right houses, and you stop spending your week on roofs that don't need you.

None of it replaces the ladder. The prediction tells you where to look; the inspection tells you what is true. Keep those two honest and separate, document what you find, leave the coverage decisions to the insurer and the claim to the homeowner, and you will out-work every competitor still treating the whole street as one undifferentiated blob.

FAQ

Can you really tell a roof's age from a satellite image?

You can estimate it as a range, not a date. Aerial imagery shows color fade, streaking, and surface texture, and comparing images across years reveals when a roof was re-roofed or how fast it is weathering. Combined with build year as a ceiling and any permit records, that triangulation puts a defensible range on the roof, like 17 to 22 years. Anyone claiming an exact installation date from imagery is using false precision.

Why isn't the year the house was built the same as the roof's age?

Because re-roofs are invisible to public records. Tax records, Zillow, and county GIS all show year built, but a 1995 house may be on its second or third roof and a 2014 house may already have a hail replacement. Build year is the oldest the roof could be, never the age it actually is. Treating the two as the same is the most common targeting mistake roofers make.

What's wrong with just using a hail map to target after a storm?

A hail map shows where it hailed, not which roofs got worn out. The swath is huge and crude, and inside it the actual damage is uneven block to block depending on hailstone size, storm direction, roof orientation and pitch, and shingle age. Knocking a whole ZIP wastes payroll on roofs the storm never really hit. A per-roof read that accounts for orientation and age ranks the doors inside the swath instead of treating them all the same.

What can I tell about a roof from the ground without a ladder?

A lot, if your eye is trained. From the curb you can read color fade, granule loss and sheen, curling or clawing shingles, missing or lifted tabs, waviness, a sagging ridge, granule piles at downspouts, and obvious flashing or pipe-boot failures. What you cannot reliably see is hail bruising, the back slopes, or the decking and ventilation. The curb read is a triage tool that tells you which roofs deserve a ladder, not a substitute for the inspection.

How do I train a new canvasser to grade roofs from the street?

Build a calibration set. Photograph ten to fifteen roofs you have already gotten on and graded by hand, label each with its real age and condition, and have the new rep grade them from the street before you show the ground truth. Two afternoons of that calibrates the eye far faster than instinct. Without a known-answer set, ground grading is just opinion and ten reps will give you ten different answers.

What signals mean an asphalt shingle roof is near the end of its life?

Heavy granule loss to the point the roof looks shiny or slick where the asphalt shows through, widespread color fade, curling or cupping or clawing shingles, missing tabs, granule piles in gutters and at downspout outlets, and cracked pipe boots or failing flashing. On architectural shingles these typically start appearing around 18-plus years in a hot or storm-prone climate, well short of the warranty number on the wrapper.

How do I turn all these signals into one decision for a street?

Build a simple weighted 0 to 100 score. Weight estimated roof age most heavily, then storm wear modeled on the roof's orientation, then the curb-condition read, then records flags like an old build year with no re-roof permit. A confirmed recent re-roof should sink a door regardless of everything else. Houses scoring high are knock-first, the middle band is a second pass, and low scores get skipped. Then feed which scored doors actually became jobs back into the weights.

Does predicting damage mean I can tell a homeowner they have a claim?

No. Prediction gives you odds, not proof. A storm model tells you a roof was likely worn or hit, which is a reason to get up there and document conditions. It is not evidence of damage and it is not a claim. The roofer documents what the inspection finds and writes an estimate, the insurer decides coverage, and the homeowner owns the claim. Telling a homeowner they have damage before you have inspected is a promise the roof has to keep.

How does a tool like RoofPredict compare to EagleView or Roofr?

They answer different questions. EagleView, Roofr, and HOVER measure a roof you already chose to bid, giving you accurate dimensions for a takeoff. RoofPredict comes earlier in the process and answers which house to go after, scoring roofs in an area by estimated age range and per-roof storm wear so you knock and mail the right doors. One tells you which roof; the other measures the roof. Most growing crews end up using both at different stages.

How often should I re-score my neighborhoods and my old CRM list?

Roofs age and storms keep coming, so a prediction is a snapshot. Re-score target neighborhoods at least seasonally, and immediately after any significant hail or wind event in your area. Re-score your CRM of old estimates a couple of times a year, because homeowners you quoted three or four years ago may have aged into the replacement window, and that book is money already sitting in your records.