How RoofPredict turns roof information into a useful plan

Coverage and data availability

RoofPredict is expanding across U.S. markets. The depth of a report varies by address because image cadence, resolution, permit access, and machine-readable property records vary by county. A well-covered property may support a narrow age window and a detailed history; another may support only a broad estimate and a list of records or field observations that would improve it.

Here is the scale, precisely. Our sealed imagery catalog holds 14,050,008 roof images covering 2,071,744 U.S. homes. 710,381 of those homes have been photographed on two or more distinct dates and 6.17 million of the images carry a capture date precise to the day, which is what makes a before-and-after read on a single roof possible at all.

• 14,050,008

  roof images in the catalog

• 2,071,744

  U.S. homes covered by those images

• 710,381

  of those homes captured on two or more distinct dates

• 6.17 million

  images carrying a capture date precise to the day

That catalog exists to train and validate our roof-dating models across multiple U.S. metros and multiple years, and it grows with every market we scan. It is not a sample of American housing: it was built where imagery is dense, coverage is uneven from county to county, and its counts describe the catalog rather than the national roof stock. We publish the exact counts so you can weigh what our estimates rest on, rather than take “millions of parcels” on faith.

The professional network also grows market by market. A match considers credentials, service area, roof type, and the reason for the visit. When available information cannot support a conclusion or the network does not yet cover a property, that limitation stays part of the answer.

What data we use

Historical roof imagery

Successive image captures of the same rooftop over time. When the surface changes color, texture, or pattern between two captures, the roof was very likely replaced in that window. The width of that window sets the width of our estimate.

Public property and permit records

County and municipal records, where available and machine-readable, including build year and roofing permits. A permit is the strongest single piece of evidence we can find without visiting the house.

Recorded weather history

Publicly recorded hail and wind events for the location, sourced from NOAA's storm reporting. This tells us what a roof has been exposed to. It does not tell us whether that roof was damaged.

Neighborhood and construction context

Homes built and re-roofed in cohorts age in cohorts. Where a subdivision was built in one cycle, that context narrows an otherwise wide estimate.

How an estimate is built

1. Find the strongest available evidence of the last replacement — a permit record if one exists, otherwise a visible change between two imagery captures.
2. Express that evidence as a window rather than a point. Two captures three years apart produce a three-year window, not a single year.
3. Widen or narrow the window using context: neighborhood build cohort, property records, and image quality.
4. Layer recorded weather exposure separately, as context about what the roof has lived through — never as a claim that damage occurred.
5. Hand the whole thing, including the uncertainty, to the inspecting professional. On-site observation overrides our estimate every time.

Residential and commercial roof underwriting models

RoofPredict is building models that help organize roof information consistently before a human makes a decision. The residential model focuses on the roof's likely age, material, replacement history, visible condition signals, recorded weather context, and the homeowner's planning horizon. The commercial model must also track roof sections, low-slope assemblies, drainage, rooftop equipment, repairs, warranties, inspection logs, and the operational consequences of a leak or replacement.

The intended output is an evidence trail, not an unexplained score: what is known, what is estimated, how strong the evidence is, what could change the conclusion, and which questions require physical confirmation. That structure can help homeowners plan, insurers focus prevention and inspection work, lenders understand a collateral question, and roofing professionals arrive with a better-prepared scope. Each party still applies its own rules and professional judgment.

How the professional network is reviewed

A network introduction is meant to focus the next step, not manufacture urgency. RoofPredict reviews the following areas before assigning a participating professional and continues to consider homeowner feedback after introductions:

• Identity, business standing, and required state or local roofing credentials are reviewed
• Insurance documentation is checked against the requirements that apply to the business and market
• Public reputation, business history, and homeowner feedback are considered
• Service area, roof type, and the reason for the visit are matched before an introduction
• Participating professionals agree to clear communication and no-pressure homeowner visits

Screening cannot guarantee the outcome of a future project. Homeowners should still verify current credentials, compare written scopes, read the contract, and decide whether the company fits their needs.

What this cannot tell you

• We produce a range, not a date. Anyone quoting your roof's exact age without a permit or an invoice is guessing.
• Imagery has gaps. Coverage frequency, resolution, tree cover, and capture season vary widely by location, and thin coverage means a wider range or no estimate at all.
• A partial re-roof can read as a full replacement. Slope-by-slope repairs, or a replacement of only the storm-facing side, can look like a whole new roof from above.
• Recorded storm exposure is not damage. A hail report near your address means hail was reported nearby, not that your shingles were bruised.
• Nothing here evaluates what is under the shingles. Decking, underlayment, ventilation, and flashing condition require someone on site.
• We do not assess insurance coverage or claim eligibility. RoofPredict is not an insurance adjuster or a public adjuster, and we do not file or negotiate claims.

Editorial standards

Our homeowner guides are written by the RoofPredict Research Team and reviewed against these rules before publication.

• Every number in a guide is either sourced to a named organization or explicitly hedged as an estimate.
• We link the sources we rely on, primary wherever they exist — trade associations, federal agencies, research institutes. Where no primary dataset is published, as with residential cost figures, we cite the named industry surveys we actually used and label them as surveys.
• We do not use fear as a conversion tactic. "Your roof is probably fine" is a legitimate conclusion and we publish it.
• Guides carry a visible last-updated date. When a figure changes, we update the guide instead of quietly leaving the old one up.
• We disclose the business model on the page: RoofPredict may receive a referral or marketing fee when a homeowner chooses to work with a network company; homeowners pay nothing for the Roof Report.
• Corrections are made in place, and material corrections are noted in the guide.

Found something wrong? Tell us at info@roofpredict.com and we will correct it.