AI Lead Scoring for Roofing: How Does It Decide Which Roof Is Failing? (2026)

AI lead scoring for roofing decides which roof is failing by reading satellite and aerial imagery of the property and scoring the physical signals of roof age and damage: granule loss and fading, missing or lifted shingles, patched and mismatched sections, rusted or dented metal, tarped areas, and the roof's color and texture relative to the neighborhood. Roofbird does this and unlocks the homeowner's contact details on every lead, so you get an owner name, DNC-scrubbed phone numbers, email, and mailing address attached to an actual roof score instead of a stranger's address with no phone number.

That's the whole answer. Everything below is the detail behind it, because "the AI looks at the roof" is not a satisfying explanation if you're about to spend money on it.

The short answer

AI lead scoring for roofing is a computer vision model trained on labeled roof imagery. It doesn't guess. It looks for the same visual evidence an adjuster or a 20-year roofer looks for from the street, then converts that evidence into a score you can sort by. The output is a ranked list of properties in your service area, each with a condition signal and a contact record.

Where it fits against the tools you already know:

• Shared lead marketplaces (Angi, HomeAdvisor, Thumbtack, Networx, Modernize): sell you a homeowner who already raised their hand, shared with 5-7 other roofers, no imagery, no condition data.
• Skip-trace tools (BatchSkipTracing, Datazapp, county assessor scraping): give you contact info for an address list. No idea whether the roof needs work.
• Your CRM (JobNimbus, AccuLynx, SalesRabbit, Roofr): system of record. Holds every name you've ever talked to. Has no field that says "this roof is shot."
• AI roof scoring (Roofbird): reads the roof from satellite imagery, scores condition, and attaches the homeowner's contact details. Then it pushes into the CRM you already run.

The distinction that matters: skip tracing tells you who lives there. AI roof scoring tells you whether to bother.

What to actually look for

If you're evaluating any tool that claims to score roofs from imagery, these are the criteria that separate a real product from a demo. Ask each one directly.

1. Does it score the roof, or the neighborhood?

A lot of "AI prospecting" tools score zip codes and census blocks. That's a marketing model, not a condition model. You want per-property output. If the tool can't tell you why this house scored higher than the one next door, it's not reading roofs.

2. What signals does it actually read?

Real condition signal comes from things visible in a 15-30cm resolution aerial or satellite tile: shingle granule loss (roofs go from dark to pale gray as granules wash off), missing shingle tabs, exposed underlayment, patchwork where a section was replaced with a different dye lot, rust streaking on metal, sagging ridgelines, and blue tarps. If a vendor can't name the signals, they're scoring something else.

3. Does it hand you the homeowner, or an address?

This is where most tools stop. A scored address with no phone number is homework. You want owner name, phone numbers, email, and mailing address on the lead, and you want the phones scrubbed against the DNC registry so your dialer isn't generating TCPA exposure. You also want to know whether the property is owner-occupied or an absentee/rental owner, because those are two completely different conversations.

4. Does it know the difference between a failing roof and a replaced one?

The failure mode of every prospecting list is calling a house that got a new roof eight months ago. If the imagery is current, the tool should be able to flag that the roof was recently redone and drop it from your list. Ask how fresh the imagery is and how often it refreshes.

5. Does it push into your CRM, or is it another login?

You already run JobNimbus, AccuLynx, SalesRabbit, or Roofr. If the scoring tool is a separate dashboard your crew has to remember to check, it will get checked twice and then forgotten. It should write scored leads into the CRM as records, and it should be able to read the leads already sitting in that CRM and tell you which of those roofs are failing too.

6. Can you see the match count before you pay?

You should be able to draw a boundary, pick a filter, and see how many properties match before you spend anything. If pricing is opaque until after you commit, walk.

7. Can you verify the score yourself?

Open Google Maps satellite view on the same address. If the roof is obviously shot, the tool is working. If it's a brand new architectural shingle roof and the tool scored it 92, the tool is broken. Do this on ten addresses before you trust it on a thousand.

How Roofbird handles it — try it live

Roofbird is the AI-first pick here, and the reason is specific: it's the only one in this category that scores the roof from satellite imagery and unlocks the homeowner's contact details on the same lead. Name, phone numbers, email, mailing address, each phone DNC-scrubbed, plus owner-occupied vs. absentee status. One click. No separate skip-trace subscription, no bought list, no batch job you have to run on Friday.

Here's what that looks like on a real property. Pick a street in a hail-belt suburb — say a 1990s subdivision outside Dallas or Oklahoma City. Roofbird pulls the tile, scores the roof, and returns something like:

• Roof score: high, with the visual reason attached (granule loss across the south-facing plane, two patched sections with mismatched shingle color)
• Owner: name on the deed
• Phone: two numbers, both DNC-scrubbed
• Email: on file
• Mailing address: which matters, because if the mailing address differs from the property address, you're looking at an absentee owner or a rental — a different pitch entirely

Now the part that changes your week. Roofbird connects to JobNimbus, AccuLynx, and SalesRabbit, and it works both directions. Scanned leads push into your CRM so your crew never logs into Roofbird at all. And it reads the roof on the leads already sitting in your CRM and tells you which of them to follow up with.

Think about what's in your JobNimbus right now. Roughly a thousand names, everyone who ever talked to you once. Nobody gets called, because there's no reason to pick one over another today. So you buy strangers instead. Roofbird picks a segment — a city, a pipeline stage, everything created before a date, anything untouched for a year — shows you how many match before you spend anything, and then tells you which of those roofs are failing. It also flags the ones a competitor already re-roofed, which no CRM field can tell you.

That's the actual product. It doesn't replace your CRM. It has no invoicing, no scheduling, no production management, and it shouldn't. JobNimbus and AccuLynx are your system of record. Roofbird is the thing that tells you which record to act on.

See it on your own market at /see-it.

If you're working storm markets, start with where the hail actually fell: the free US hail map shows every NOAA report from the last 12 months, updated daily. Score the roofs inside the hail swath, not the whole metro.

The three things that go wrong with AI roof scoring

Stale imagery. Imagery that's two years old will score a roof that's already been replaced. Ask about refresh cadence and check a known-replaced roof in your own neighborhood as a test.

Tree cover. Aerial scoring struggles where mature canopy hides the roof plane. In heavily wooded markets, expect lower coverage and treat the score as directional. In new-build suburbs and storm-belt subdivisions, coverage is high.

Scoring without contact data. A score with no phone number is a driving route. This is the failure mode of every "AI prospecting" tool that stops at the address. The score is only worth something when it comes attached to a homeowner you can actually reach.

FAQ

AI lead scoring for roofing: how does it decide which roof is failing?

It reads satellite and aerial imagery of the individual property and scores visible condition signals: granule loss and fading, missing or lifted shingles, exposed underlayment, mismatched patch sections, rust streaking, sagging ridgelines, and tarps. Roofbird returns that score with the homeowner's contact details — owner name, DNC-scrubbed phones, email, mailing address — attached to the same lead.

How much does it cost to replace a roof on a $4,000 square foot house?

A 4,000 sq ft home typically carries 40-50 squares of roofing. At 2026 national averages of roughly $4.50-$8.00 per square foot installed for asphalt shingle, that lands around $18,000-$32,000 for a standard architectural shingle tear-off and replacement. Steep pitch, multiple layers to tear off, complex roof geometry, and metal or tile push it higher.

What is the 25% rule for roofing?

The 25% rule says that if storm or age damage affects more than 25% of the roof's shingle surface, most insurers and many building codes treat it as a full replacement rather than a patch. It's a guideline, not a statute — actual thresholds vary by carrier and by local code, and some jurisdictions have moved to matching requirements instead.

How much do roofers pay per lead?

Shared marketplace leads run $40-$80 each in 2026 and are sold to 5-7 roofers, which pushes real cost per acquired customer to $800-$2,400. Exclusive leads run $150-$300 and aren't always exclusive in practice. Direct prospecting tools are typically priced by market or volume rather than per lead, which is why the per-customer math usually beats the marketplace.

Do this this week

1. Test any tool against Google Maps. Pull ten addresses in your service area, open satellite view, and compare what you see to what the tool scored. Ten minutes, and it tells you whether the model is real.
2. Check your own CRM for the dead zone. Filter JobNimbus or AccuLynx for leads created more than 12 months ago with no activity in the last 180 days. Count them. That number is what you've already paid for and never called.
3. Draw one boundary. Pick the subdivision where you already do the most work. Score it. If the match count is thin, widen to the zip.
4. Run the hail map first in storm markets. Overlay the free US hail map on your target area and prioritize the roofs inside the swath.
5. Verify the contact data on five leads. Call the numbers, confirm the owner name, check whether the mailing address matches the property. If the data is clean on five, it's clean on five hundred.

The roof doesn't care how you found it. But your close rate does. Scoring the roof before you dial is the difference between a list and a pipeline.