How to Prioritize Roof Inspections by Likelihood of Replacement

Most roofing companies inspect roofs in the order the calendar hands them out. A canvasser knocks a street top to bottom. A storm rolls through and everyone races the same ZIP. A lead comes in, somebody climbs the ladder, and three hours later they write up a roof that had eight good years left. None of that is a strategy. It is motion that feels like progress.

The roofs that actually get replaced are a small fraction of the roofs you could inspect. On a given residential street, the homes due for a tear-off are not evenly spread, and they are almost never the ones that look worst from the curb. A roof can be filthy, mossy, and ugly and still have a decade of service life. A roof can look clean and tight and be two years past the point where an adjuster will pay to replace it. Your eyes at street level are a weak signal. The goal of prioritizing inspections is to put your most expensive resource — a trained person on a ladder — only on the roofs where the odds of a replacement-grade outcome are high enough to justify the climb.

This is a ranking problem, not a coverage problem. You are not trying to inspect every roof. You are trying to order the roofs so the ones most likely to convert to a replacement get climbed first, get climbed by your best people, and get climbed before a competitor beats you there. Below is the full system: the signals that actually predict replacement, how to weight them, the scoring model, the field workflow, the documentation that protects the job, the legal lines you cannot cross, and the metrics that tell you whether your prioritization is working.

Why "inspect everything" quietly bankrupts a crew

Start with the unit economics, because they are the whole argument. A real roof inspection — drive time, the ladder set, walking the field, photographing slopes, measuring or confirming measurements, and writing it up — costs a company somewhere between 45 minutes and 90 minutes of a paid person's time, plus fuel and vehicle wear, plus the opportunity cost of the inspection they did not do instead. Treat the loaded cost of an inspector-hour as real money, because it is. Whatever your number is, multiply it across a week of climbing roofs that were never going to convert and you can see the leak.

The other half of the cost is hidden. Every roof your crew climbs that does not convert is a roof where you spent your credibility. You knocked, you asked for access, you got on the homeowner's property, and you walked away with nothing — and now that homeowner has been "inspected" and is slightly more resistant to the next honest contractor. Burning through low-odds roofs wastes more than hours. It salts the field.

So the question is not "how do we inspect more roofs." It is "how do we inspect fewer roofs and replace more of them." That only happens if you can rank addresses before anyone leaves the truck.

What a roofer's eye gets wrong

Pros over-trust three street-level signals and under-trust the ones that matter.

• Curb appearance. Granule loss, streaking, and moss are visible from the ground, but they correlate poorly with whether a roof is replaceable on a storm claim or aged out. Algae streaking (the black stains) is largely cosmetic. A roof can look terrible and pass; a roof can look fine and fail.
• Visible sag or obvious failure. By the time you can see deck sag or active leaks from the street, you are usually past the inspection-prioritization stage and into emergency work. Those are real, but they are rare and self-announcing.
• The neighbor's new roof. "The house next door got replaced, so this one's due" is a decent tiebreaker and a terrible primary signal. Re-roofs happen for a hundred idiosyncratic reasons — a sale, a leak, an insurance push — that say nothing about the house beside it.

The signals that actually predict a replacement are roof age, storm exposure modeled to the specific roof, the material and its real-world service life, slope and complexity, and a handful of condition cues that you can only partly read from the ground. The rest of the work is turning those into a number you can sort by.

The five signals that predict a replacement

Think of every address as carrying a small set of measurable signals. You will combine them into one score, but you have to understand each one on its own first, including what it can and cannot tell you.

1. Roof age (as a range, never a date)

Age is the single strongest predictor of replacement likelihood, and it is the one most companies estimate worst. People reach for the year the house was built — Zillow, the county assessor, Google — and that number is almost useless for roofing. The year built tells you when the original roof went on. It says nothing about the re-roof that happened in year 19, or the storm replacement in year 11, or the flip that put a fresh roof on before sale. A 1994 house can have a 2021 roof. Tax records and listing sites do not see re-roofs.

What you actually want is the age of the current roof, and the honest form of that answer is a range, not a date. From aerial and historical imagery you can often bracket a roof: it was present and looked aged in one image set, looked newer in an earlier one, and the transition narrows the window. The output you should trust and act on looks like "this roof is roughly 17 to 22 years old," not "this roof was installed on March 3, 2003." Anyone selling you an exact install date from imagery is overselling. A tight range is enough to rank.

Why age dominates: most asphalt roofs are replaced not because of a single event but because they reached the back third of their service life and then a storm, a leak, or a sale pushed them over. A roof in years 5 to 12 rarely converts to a full replacement even after moderate hail, because there is too much life left for a replacement-grade outcome. A roof in years 18 to 28 converts far more readily, because the same hail event lands on a roof that is already brittle, already losing granules, already a candidate. Age does more than stand alone — it multiplies every other signal.

2. Storm exposure, modeled to the roof, not the ZIP

The second signal is what the weather actually did to this roof. Most contractors work off a hail map — a polygon showing where hail of some size was reported or radar-estimated. That tells you a storm passed through an area. It does not tell you which roofs it wore out.

The gap between those two things is enormous. Hail size, fall direction, wind speed and direction at the moment of impact, the slope orientation of a given roof, the material, and the roof's age all determine whether a specific roof took functional damage or shrugged it off. Two houses on the same street, under the same storm, can have completely different outcomes: the one whose steep south slope faced into wind-driven hail gets bruised across the field; the one tucked behind a treeline with shallow slopes facing away barely registers. A ZIP-level hail polygon treats them as identical. They are not.

This is why "storm exposure" as a prioritization signal has to be modeled per roof, combining hail and wind physics with the roof's own geometry and age, rather than read off a map. The honest framing for any storm signal is odds, not proof. Modeling tells you which roofs most likely took replacement-grade damage. It does not certify damage — only a physical inspection and, ultimately, an adjuster's call does that. But for the narrow job of ranking which roofs to climb first after a storm, per-roof storm modeling is dramatically better than a polygon, because it separates the worn-out roofs from the merely rained-on ones.

3. Material and its real service life

The material sets the clock. A signal that looks alarming on one material is routine on another.

The practical takeaways: asphalt — especially 3-tab — is where age and storm signals convert into replacements most often, which is why most volume residential operations weight asphalt roofs higher in the queue. Tile and metal roofs throw a lot of false positives; an "old" tile roof is often nowhere near the end of its life, and a dented metal roof is frequently a cosmetic-only argument that does not reach replacement. You do not have to ignore those roofs, but they should carry a lower base score unless something specific elevates them.

You usually cannot identify material with certainty from overhead imagery alone — asphalt vs. some laminate products, certain tile profiles, and metal can be ambiguous from above. Treat material as a probable classification that the inspector confirms on site, and weight your confidence accordingly.

4. Slope, complexity, and geometry

Geometry feeds the score in two ways.

First, damage susceptibility. Steeper slopes facing the storm's approach take more direct hail impacts. Complex roofs — lots of valleys, hips, dormers, penetrations, and transitions — accumulate more wear and more storm-vulnerable detail per square. A simple gable ranch and a cut-up multi-story with six dormers age very differently even at the same calendar age.

Second, job economics, which belong in your prioritization even though they are not about damage. A 12/12 cut-up roof with three stories of access problems is a harder, slower, more dangerous inspection and a more expensive job. If two roofs score equally on replacement likelihood, the one your crew can inspect and produce safely and quickly is the better use of the next ladder set. Prioritization is about return on the inspection hour, and complexity is part of that return.

5. Condition cues you can partially pre-read

Some condition signals show up from the ground or from imagery and can nudge a score before the climb: obvious missing shingles or tarps, prior patchwork that signals an owner nursing a failing roof, debris and tree overhang that accelerate wear, and visible storm damage to soft metals nearby (gutters, downspouts, screens, AC fins) that hints the area took real hail. These are supporting signals, not primary ones. Use them to break ties and to flag a roof for an earlier climb, not to carry the ranking by themselves.

Turning signals into a score you can sort by

Now assemble the signals into a single number. The point of a score is not false precision — it is a consistent, repeatable ordering so the queue is the same no matter which manager built it or which canvasser worked it. A simple weighted model beats gut feel because it is consistent and you can tune it against your own close data.

Here is a starting framework. Score each address 0-100; tune the weights to your market after you have outcome data.

A few rules that keep the model honest:

• Age and storm are multiplicative in reality, not merely additive. A storm on a new roof is a low-conversion event; the same storm on a worn roof is a high-conversion one. If you want one refinement beyond the additive table, multiply the storm component by an age factor (e.g., 0.3 for new roofs, 1.0 for back-third roofs) so a big hail event on a young roof cannot float to the top of the queue on storm points alone.
• Confidence is its own dimension. A roof scoring 80 from confirmed signals is not the same as a roof scoring 80 where material is a guess and the storm read is ZIP-level. Carry a confidence flag (high / medium / low) alongside the score. High-score, high-confidence roofs go to your best closers first.
• Never let one signal alone push a roof to the top. A roof that scores high only on age (no storm, ambiguous material) and a roof that scores high only on a fresh storm (but is a 4-year-old architectural roof) are both lower-conviction than a roof that lights up on two or more signals. Reward agreement between signals.

A worked example

Three addresses on the same block after a spring hail event.

123 Maple — built 1996, asphalt 3-tab, simple gable, modeled storm odds high. Age read from imagery: 16-21 years on the current roof (a re-roof is visible in older imagery, so it is not the 1996 original). That is squarely back-third for 3-tab. Age scores ~32/35. Per-roof storm model says the south slope faced wind-driven 1.25-inch hail: ~27/30. Material 3-tab: 15/15. Simple roof, easy access, storm-facing slope: 8/10. Soft-metal denting visible on the gutters from the street: 8/10. Total ≈ 90, high confidence. Climb this first, send your best person.

127 Maple — built 2014, architectural asphalt, moderate complexity, same storm. Roof is the 2014 original, roughly 9-12 years old: first-to-mid life. Age scores ~10/35. Storm model: same hail, but the roof is young and the age multiplier knocks the storm component down to ~12/30. Material architectural: 11/15. Some complexity, decent access: 6/10. No pre-readable damage: 3/10. Total ≈ 42, medium confidence. This is a maybe — worth a knock to leave a card, not worth bumping ahead of Maple 123. A young architectural roof rarely converts even under real hail.

131 Maple — built 1988, concrete tile, complex roof, same storm. Tile roof, hard to age from imagery, possibly 20-30 years on the underlayment but the tile itself reads as having decades left. Age signal for replacement is weak — tile rarely converts on age. Storm model: tile sheds hail far better than asphalt, modeled odds low. Material tile: 4/15. Complex, three stories, brutal access. Total ≈ 30, low confidence. Deprioritize. The cut-up tile roof is the kind of address that eats two hours and produces nothing.

Notice the result: the roof a tired canvasser might skip (the plain little ranch) outranks the big impressive house, and the model says so before anyone climbs. That is the entire point.

How to actually bracket a roof's age from imagery

Since age carries the most weight, it is worth being concrete about how you turn pictures into a usable range. You are looking for transitions across time. Modern imagery providers keep historical captures of the same property going back years; comparing them lets you find the window where a roof changed.

The tells that a roof is aged in a given image: uneven, blotchy color across slopes; visible granule-loss patterning (lighter and darker zones as the mat shows through); curling or cupping at the edges casting small shadows; patched sections that don't match the surrounding field; and streaking that has had years to spread. The tells that a roof is recent: uniform color, crisp edges, sharp shadow lines at ridges and hips, clean valleys, and no patchwork. When you find an image where a roof reads aged and an earlier one where the same roof reads fresh, the replacement happened between those two dates — and the more captures you have, the tighter you can squeeze the window.

Three honest cautions. First, sun angle and image quality fool the eye; a low-sun capture exaggerates texture and can make a fine roof look worn. Cross-check more than one image before you trust a read. Second, color alone is weak on certain materials — weathered architectural shingle and a deliberately dark product can look similar from above. Third, you are bracketing the current roof, so if a re-roof is visible in the timeline, your range starts from that re-roof, not from the original construction. The output you carry into the score is a range with a confidence level attached — "14 to 19 years, high confidence" reads differently in the queue than "roughly 15 to 25 years, low confidence," and your model should treat them differently.

Doing this by hand for one roof you are standing in front of is fine. Doing it for a thousand roofs across a neighborhood is exactly the kind of work that does not scale by hand, which is the practical reason most companies never built an age-ranked queue until the signal could be produced at area scale.

Building the inspection queue: a step-by-step workflow

Scoring produces a number. A queue turns the number into a day's work. Here is the operational loop.

1. Define the working area. A neighborhood, a set of streets, or the footprint of a recent storm. Keep it tight enough that drive time between climbs is minutes, not tens of minutes. Clustered routes beat scattered high scores; a 92 across town can lose to three 80s on one street once you price the windshield time.
2. Pull or build the signals for every address in the area. Age range, per-roof storm read, probable material, geometry, and any pre-readable condition cues. This is the step that used to be impossible to do at scale and is now the part you can buy or generate instead of guessing house by house.
3. Score and sort. Apply the weighted model. Produce a ranked list with the score and the confidence flag visible.
4. Set thresholds, not only an order. Decide the cutoffs: high-conviction roofs (say 75+ with medium-or-better confidence) get a full inspection push; mid roofs (50-74) get a knock and a card but not a guaranteed climb; low roofs (under 50) get skipped or mailed, not climbed. The threshold is where the queue saves you money — it tells the crew where to stop, not only where to start.
5. Route the high-conviction roofs into clusters. Hand each canvasser or inspector a tight geographic run of high-score addresses, not a spreadsheet of the whole town. The route should minimize drive time while front-loading the best roofs early in the day, when your people are sharp.
6. Capture the outcome of every climb. Replaced / repairable / no-action, and why. This is the data that lets you tune the weights. Without outcome capture you are guessing forever.
7. Re-score after the day and after the next storm. A roof that was a 60 last month is a 90 the day after a hail event lands on it. The queue is not static; it breathes with the weather and with age.

Don't forget the roofs already in your book

The queue does not have to be built only from cold streets. The cheapest high-conviction roofs you will ever inspect are often sitting in your own records: estimates you wrote that never closed, repairs you did years ago, and past customers whose roofs have quietly aged since you last spoke. A roof you bid as a borderline repair seven years ago may now be a back-third replacement candidate, and you already have the address, the owner's name, and a prior relationship — which converts far better than a cold knock.

The move is to run your CRM and old estimate list through the same scoring model. Pull the roof-age range and storm read for every address you have already touched, and surface the ones that have crossed into replacement-likely territory since the last contact. These roofs carry a built-in trust advantage and zero acquisition cost, and they are routinely the highest return-on-inspection-hour addresses a company has — money already sitting in the book, waiting for someone to re-score it. Most companies never do this, because re-scoring a five-year-old customer list by current roof age is, again, work that does not happen by hand.

Choosing thresholds for your market

The right cutoff depends on how much inspection capacity you have versus how many roofs you can mail or skip. If your crew is the bottleneck (too many roofs, not enough ladders), raise the threshold so they only climb the highest-conviction roofs and you mail the rest. If your area is the bottleneck (you have crew capacity but few roofs in the queue), lower the threshold and climb deeper into the mid-tier. Prioritization is a throttle, not a fixed line — set it against your actual constraint.

Where roof-due data comes from (and where RoofPredict fits)

Everything above assumes you can get the signals for every address before you knock. For years that was the wall: you could score a roof you were standing in front of, but you could not score a thousand roofs from your desk. So companies fell back to inspecting in calendar order. The pieces that close that gap come from a few different sources, and it helps to know what each one actually gives you.

• Listing sites and county records (Zillow, the assessor) give you year built, which — as covered above — is the wrong number. They do not see re-roofs. Useful for owner-occupancy and basic property data; not for roof age.
• Measurement vendors (EagleView, HOVER, and similar) give you precise roof measurements — squares, pitch, facets, lengths — which you need to estimate and produce a job. They answer "how big and what shape is this roof," not "which roof is due." Different category. You use them after you have decided a roof is worth pursuing, to build the estimate.
• Hail maps and storm-data vendors give you area-level storm history — where hail of a given size was reported or estimated. Good for confirming a storm happened; weak for telling you which specific roofs it wore out, because they are polygons, not per-roof physics.

What the scoring model needs and those sources do not fully provide is the combination: a per-address roof-age range plus storm modeled to that specific roof, across every house in an area, so you can rank before you climb. That is the specific gap RoofPredict is built to fill. It takes aerial imagery and weather data and returns, house by house, a roof-age range, a per-roof storm read (hail and wind modeled to the roof's geometry and age, not a ZIP polygon), and a resulting risk score — the exact inputs the prioritization model above runs on. The honest framing matters: the age comes back as a range, and the storm read is odds, not proof — a ranking signal that tells you which roofs most likely took replacement-grade wear, not a certification of damage. It is not a lead service and it does not climb the roof for you. It enriches your own list — your streets, your CRM, your mailing area — with the roof-age and storm signals so you can sort it and skip the roofs that are not due.

That last part is where it pays off twice. It is just as valuable for the roofs you skip as the ones you climb. Pulling the new roofs and the young architectural roofs out of the queue before anyone leaves the truck is where a crew stops salting the field on low-odds doors. And it works on a book you already own: feed it your old estimates and past customers and it tells you which of those roofs have aged into the back third since you last talked to them — money already sitting in your CRM. The limits are real and worth saying plainly: an age range is not an install date, a storm model is not an inspection, and material read from above is a probable classification the inspector still confirms on the roof. Used for what it is — a way to rank a thousand roofs before you climb the right fifty — it turns "inspect in calendar order" into "inspect in likelihood order," which is the entire game.

Documenting the inspection so the job survives

Prioritization gets your best person onto the right roof. What happens on that roof determines whether the job holds together. The single biggest mistake in storm and replacement work is a thin inspection — a few phone photos, a vague write-up, and a homeowner left to fend for themselves. Thorough documentation is what separates a roof that gets a replacement-grade outcome from one that stalls.

Here is the documentation standard for a high-conviction inspection.

The photo set

• Establishing shots. The full house from the street, all four elevations, and a wide shot of the roof from a corner that shows the overall plane. These orient anyone who later reviews the file.
• Every slope, full field. Walk each slope and photograph the field, not only the spots you think look bad. You are documenting the whole roof's condition.
• Damage close-ups with scale and location. Each impact, crease, or failure photographed close, with a chalk circle or a reference object for scale, and a wider shot that places it on the roof so it is clear where it is. A close-up with no context is a weak photo.
• Soft metals and accessories. Gutters, downspouts, vents, flashing, drip edge, valleys, AC condenser fins, window screens, and any nearby soft-metal surfaces. Hail leaves dents in soft metals at the same event that bruised the shingles; those collateral hits corroborate the date and severity.
• The penetrations and transitions. Pipe boots, step flashing, chimney and skylight details, valley conditions. These are where roofs actually fail and where condition is most honest.
• Test squares where appropriate. A marked square (commonly a 10x10 area) with the impacts within it circled, so density of damage is documented in a defined area rather than asserted.

Date and geotag everything. A documented inspection is a dated, located, complete photographic record of the roof's condition — that is its job.

The write-up

The written inspection should describe what you observed, slope by slope, in plain factual language: the material, the approximate age condition, the damage you found and where, the soft-metal collateral, and the scope of repair that the observed condition would require. Then build that scope into an accurate estimate.

This is the line you have to understand cold: your job is to document thoroughly and write an accurate, Xactimate-aligned estimate to repair the roof, and hand it to the homeowner. Aligning your estimate to the same line items and pricing structure adjusters use makes it easy to read against the carrier's own numbers and keeps the conversation factual. The homeowner files; the insurer decides coverage. You are the documentation-and-estimate side of that, and only that.

The legal lines you cannot cross

This matters enough to be its own section, because the same prioritization work that finds storm-worn roofs walks you straight up to the edge of unlicensed public adjusting, and crossing that line gets companies fined and shut down. Most states draw a hard distinction between a contractor working on their own scope and a public adjuster representing the homeowner against the carrier for a fee. Know which side you are on.

What a roofing contractor may do: inspect the roof, document the damage thoroughly, prepare an accurate estimate to repair their own work/scope, state facts about that scope to the carrier, and hand the homeowner a clear, professional document package. Facts about your scope. Documentation. An estimate. That is your lane, and it is a powerful one.

What a roofing contractor may not do — the do-not-say list:

• Do not, for a fee, negotiate, adjust, or "handle" the homeowner's claim. That is public adjusting and it is licensed in most states.
• Do not interpret the policy or tell the homeowner what is or is not covered. Coverage is the insurer's call and a coverage opinion is adjusting.
• Do not promise a specific payout, approval, or outcome. You document odds and facts, not guarantees.
• Do not promise the deductible will be waived, absorbed, eaten, or "taken care of." Offering to cover a deductible is illegal in many states and is insurance fraud framing everywhere. Say nothing about the deductible except that the homeowner is responsible for it.
• Do not advertise or imply a "free roof." The roof is not free; a claim may or may not pay, the homeowner owes their deductible, and "free roof" marketing is exactly what regulators target.
• Do not represent the homeowner against the insurer. You represent your own scope and documentation. The homeowner deals with their carrier.

The safe frame, said simply: document thoroughly, write an accurate estimate, hand it to the homeowner — the homeowner files and the insurer decides. Teach this to your canvassers and inspectors as a script, because a green rep trying to be helpful is exactly who blurts out "we'll get your deductible waived" or "this'll definitely be approved" and creates a liability the company carries. Train the do-not-say list as hard as you train ladder safety.

Safety and access: prioritization protects your people too

There is a safety argument for prioritization that owners underrate. Falls are the leading cause of death in construction, and roofing carries one of the highest fatal-fall rates of any trade. Every roof you climb is exposure. Climbing roofs that were never going to convert is worse than wasted money — it is unnecessary risk taken for no return.

Prioritization reduces that exposure directly. Fewer climbs means fewer ladder sets, fewer people on steep and complex roofs, and fewer high-risk inspections done on low-odds houses. When complexity and access difficulty are part of your scoring (as recommended above), you are also steering your crew away from the most dangerous climbs unless the replacement odds genuinely justify them. Fall protection, proper ladder setup, and trained crews are non-negotiable on every climb — and the best fall protection is not climbing a roof you had no business climbing in the first place. Honor OSHA's fall-protection requirements on every roof, and let the score keep your people off the ones that were never worth the risk.

What pros get wrong about prioritization

A list of the mistakes that quietly wreck a well-intentioned system.

• Treating year built as roof age. Covered above, and it is the number-one error. Re-roofs are invisible to tax and listing data. If your queue is sorted by year built, it is sorted by the wrong number.
• Working a street top to bottom. Geographic order is not likelihood order. The house with the worn roof might be the third one, not the first, and a top-to-bottom knock burns your sharpest hours on whatever happens to be at the corner.
• Chasing the storm polygon instead of the roof. A hail map gets the whole crew into the same ZIP, where they fight each other for roofs that mostly did not take functional damage. Per-roof modeling separates the worn roofs from the rained-on ones inside that polygon.
• Over-weighting curb appearance. The ugliest roof on the street is frequently not the most replaceable one, and the cleanest roof can be the most due. Eyes at street level are a weak signal; trust the model over the gut on appearance.
• No confidence flag. A score with no confidence attached sends your best closer to a roof where the material was a guess. Separate "high score" from "high score we are sure about."
• No outcome capture. Companies build a scoring model, run it for a season, and never feed the results back in. Your close data is the only thing that tunes the weights to your actual market. Capture replaced / repairable / no-action on every climb or the model never gets smarter.
• Letting a green rep freelance the claims conversation. The do-not-say list is not optional. One "we'll cover your deductible" can cost more than a month of jobs.
• Ignoring job economics in the score. Two equal-likelihood roofs are not equal if one is a simple gable and the other is a three-story cut-up. Return on the inspection hour includes how hard the roof is to inspect and produce.

Measuring whether your prioritization is working

A prioritization system you do not measure is just a different way to guess. Track these and review them monthly.

The most important of these is score-band conversion. If your 75+ roofs do not convert noticeably better than your sub-50 roofs, the model is not working and you need to re-examine your signals and weights. If they do, you have proof the prioritization is earning its keep, and you can confidently raise the threshold and stop climbing the bottom of the queue.

A 30-day plan to install this

If you are starting from "we inspect whatever comes in," here is how to stand the system up without stopping the business.

Week 1 — Pick one area and get the signals. Choose a single neighborhood or a recent storm footprint. Get roof-age ranges, per-roof storm reads, probable material, and geometry for every address in it — built, bought, or pulled. Do not boil the ocean; one area is enough to learn on.

Week 2 — Score, set thresholds, and route. Apply the weighted model, sort, set your high/mid/low cutoffs against your crew's actual capacity, and build tight geographic routes of the high-conviction roofs. Brief the crew on the do-not-say list and the documentation standard before anyone climbs.

Week 3 — Run it and capture every outcome. Climb the high-conviction queue, mail or card the mid-tier, skip the low-tier. Record replaced / repairable / no-action and the reason for every single climb. This is the week the data starts to accumulate.

Week 4 — Measure and tune. Pull the score-band conversion numbers. If the high band is out-converting the low band, raise the threshold and expand the area. If it is not, look at which signals are firing on the roofs that did not convert and adjust the weights. Then repeat with a bigger area.

Within a season you will have a model tuned to your market, a crew that climbs the right roofs in the right order, and a queue that re-sorts itself every time a storm rolls through or a roof ages another year into the back third. You will inspect fewer roofs and replace more of them — which was always the point.

The roofs that need you are already out there, spread thin across streets you are working the hard way. Knowing which ones are due, house by house, before anyone climbs a ladder, is the difference between a crew that runs on motion and a crew that runs on odds. If you want the roof-age-range and per-roof storm signals that feed a model like this — enriching your own streets and your own customer book so you can rank before you climb — that is exactly what RoofPredict is built to hand you. Book a demo, bring a roof you already know the answer on, and judge the signal for yourself.

FAQ

What is the single best signal for predicting whether a roof will be replaced?

Roof age, expressed as a range rather than a date, is the strongest single predictor. Most asphalt roofs convert to a full replacement once they reach the back third of their service life and then a storm, leak, or sale pushes them over. Age also multiplies every other signal: the same hail event that does nothing to a five-year-old roof can finish off a twenty-year-old one. The catch is that you have to estimate the age of the current roof, not the year the house was built, because re-roofs are invisible to tax records and listing sites.

Why can't I just use the year the house was built to estimate roof age?

Year built tells you when the original roof went on, not when the current one did. Re-roofs from storms, leaks, sales, and ordinary aging are not recorded in county assessor data or on listing sites, so a 1995 house can easily have a 2020 roof. Sorting an inspection queue by year built sorts it by the wrong number. You want the age of the roof that is on the house right now, which you can bracket as a range from aerial and historical imagery.

How is per-roof storm modeling different from a hail map?

A hail map is a polygon showing where hail of some size was reported or radar-estimated across an area. It treats every roof inside the polygon as identical. Per-roof storm modeling combines hail and wind physics with a specific roof's slope orientation, geometry, material, and age to estimate whether that roof likely took replacement-grade wear. Two houses on the same street under the same storm can have opposite outcomes based on which slopes faced the wind-driven hail. The map confirms a storm happened; the per-roof model tells you which roofs it actually wore out — as odds, not proof.

How do I turn these signals into a score I can sort by?

Use a weighted 0-100 model. A solid starting point is roof age 35 points, per-roof storm exposure 30, material fit 15, slope and complexity 10, and pre-readable condition cues 10. Refine it by making the storm component multiplicative with age so a big storm on a young roof can't float to the top on storm points alone, and carry a confidence flag alongside the score so a guess-heavy 80 is treated differently from a confirmed 80. Then tune the weights against your own close data over a season.

Should I inspect tile and metal roofs the same way as asphalt?

No. Asphalt — especially 3-tab — converts on age and storm signals far more often than tile, metal, or slate. An old tile roof is frequently nowhere near the end of its life because the underlayment fails long before the tile, and dented metal is often a cosmetic-only argument that does not reach replacement. Give those materials a lower base score unless a specific signal elevates them, and remember that material read from overhead imagery is a probable classification your inspector confirms on the roof.

How many roofs should I inspect to land one replacement?

There is no universal number because it depends on your market, material mix, and how recently a storm hit — but the metric to watch is climbs per replacement, and the whole point of prioritization is to drive it down over time. If you track score-band conversion and your high-conviction roofs (say 75+) are not converting noticeably better than your low-tier roofs, the model is not working. When the high band clearly out-converts the low band, you can raise your threshold and stop climbing the bottom of the queue.

What documentation should a high-priority inspection capture?

Establishing shots of the house and all elevations, every slope photographed across the full field, damage close-ups with scale and a wider shot showing location on the roof, soft metals like gutters and AC fins that corroborate hail, penetrations and transitions where roofs actually fail, and test squares where appropriate. Date and geotag everything. Then write a factual slope-by-slope description and build it into an accurate, Xactimate-aligned repair estimate you hand to the homeowner.

Can I tell the homeowner the insurance claim will be approved or the deductible waived?

No. Promising a specific approval or payout, interpreting what the policy covers, or offering to waive, absorb, or cover the deductible all cross into unlicensed public adjusting or outright fraud framing, and they are illegal in many states. Your lane is to document thoroughly, write an accurate estimate for your own scope, state facts about that scope, and hand the package to the homeowner. The homeowner files the claim and the insurer decides coverage. Say nothing about the deductible except that the homeowner is responsible for it.

What does RoofPredict actually provide, and what are its limits?

RoofPredict takes aerial imagery and weather data and returns, house by house, a roof-age range, a per-roof storm read modeled to the roof's geometry and age, and a resulting risk score — the inputs a prioritization model runs on. It enriches your own list of streets or your CRM so you can rank before you climb and skip the roofs that aren't due. The honest limits: the age is a range, not an install date; the storm read is odds, not proof of damage; and material read from above is a probable classification your inspector confirms on the roof. It is not a lead service and it does not perform the inspection.

How does prioritizing inspections improve crew safety?

Falls are the leading cause of death in construction and roofing carries one of the highest fatal-fall rates of any trade, so every climb is real exposure. Climbing roofs that were never going to convert is unnecessary risk for no return. Prioritization reduces total climbs, and scoring complexity and access difficulty into the model steers crews away from the most dangerous roofs unless the replacement odds genuinely justify them. Fall protection and proper ladder setup are still required on every climb — the best risk reduction is simply not climbing a roof you had no business climbing.