Physical underwriting for commercial property
Price property risk on physical truth.
RAKE ML predicts commercial roof failure timing so insurers, lenders, owners, brokers, and underwriters can make asset-level decisions before the claim, covenant breach, or capital surprise.
Built from imagery, climate exposure, repair history, contractor feedback, and physics-informed simulation.
The blind spot
Roof risk is still priced from stale proxies.
Property teams make large decisions from roof age, sparse inspections, accounting records, and loss history that arrives too late. RAKE ML adds the missing layer: time-bound physical failure intelligence at the asset and portfolio level.
Claim exposure
$31B/yr
U.S. commercial roof claims paid each year for failures that can often be anticipated earlier.
Premium leakage
$1.31B/yr
Lost when roof age is misclassified and risky assets are underpriced at origination.
Prediction window
+/- 18 mo
The rough accuracy of age-based actuarial proxies. Better timing changes underwriting.
Figures reflect RAKE deck references to CIAB 2024, Verisk roof-age product data, and RAKE benchmark work.
What RAKE returns
An address in. A capital plan out.
The output is not another condition grade. It is a dated risk curve, a confidence interval, and an action plan your team can defend in underwriting, asset management, and credit review.
Sample asset
4200 Commerce Dr, Dallas
124,000 sq ft commercial roof
Model confidence
87%
- Failure probability, 12 mo
- 71%
- Predicted failure window
- M7-M11
- Recommended action
- Full replacement, Q3
- Cost if deferred 12 mo
- $1.74M
Illustrative output based on the RAKE overview deck, not a live asset report.
Buyers
Same physical engine. Different pricing surface.
A roofer sees opportunity. An owner sees CapEx timing. An insurer sees loss selection. A lender sees reserve and covenant risk.
CapEx + portfolio risk
Owners & Asset Managers
Prioritize repair-versus-replace decisions, avoid surprise capital calls, and negotiate from verified asset data.
Decision: act now, defer, sell, or insure differently.
Underwriting + loss control
Insurers & Underwriters
Replace age proxies with time-bound failure intelligence. Decline bad risks, load premium where risk is real, and exit before the claim window opens.
Decision: bind, price, inspect, non-renew, or prepare claims response.
Reserve + covenant risk
Lenders & Servicers
Monitor collateral condition against reserve tables and spot physical deterioration before it becomes a covenant issue or workout.
Decision: adjust reserves, flag collateral, or intervene early.
Placement + advisory leverage
Brokers & Risk Advisors
Bring defensible asset evidence into placement, renewal, and lender conversations instead of negotiating from incomplete roof schedules.
Decision: document the risk, improve submissions, and reduce friction.
Workflow
From physical signals to underwriting action.
RAKE ML fuses observable asset data with physics-informed simulation, then closes the loop with verified repair and replacement outcomes.
01
Ingest
Address, roof area, aerial and satellite imagery, inspection history, weather exposure, and repair records.
02
Simulate
Material degradation, thermal cycling, drainage failure, membrane mechanics, hail, UV, freeze-thaw, and precipitation.
03
Score
A causal model returns a time-bound failure probability curve with a confidence interval and audit trail.
04
Learn
Contractor repair events, work orders, and replacements recalibrate the model after real-world outcomes.
The RAKE engine
Real failures are too rare to learn from directly.
Most portfolios do not generate enough major roof failures to train a reliable model from claims alone. RAKE ML uses synthetic data generation and field outcome feedback to model how roofs age, degrade, and fail.
Physics-informed simulation
Component-level modeling of material degradation, drainage failure, membrane mechanics, and thermal stress.
Climate calibrated
Hail, UV load, freeze-thaw, wind, and extreme precipitation layered against geographic exposure.
Ground-truth feedback
Contractor work orders, repairs, and replacements sharpen the model with verified outcomes.
Audit-ready evidence
Every score is tied to data inputs, confidence, timing, and recommended action.
Why timing matters
"Bad roof" is not enough. A roof expected to fail in 8 months is an underwriting action. A roof expected to fail in 24 months is a pricing, reserve, or CapEx decision.
Questions physical-risk teams ask first
What is physical underwriting?
Physical underwriting prices risk from the condition and failure trajectory of the asset itself, not only from age, loss history, replacement cost, or financial records. RAKE ML starts with commercial roofs because roof failure is expensive, visible, modelable, and chronically under-measured.
What data does RAKE ML use?
We fuse aerial and satellite imagery, portfolio data, inspection history, weather exposure, repair records, and contractor outcome feedback. Available inputs vary by portfolio, so reports show confidence alongside the risk signal.
How is this different from a roof-age model?
Roof age is a proxy. RAKE ML estimates how a roof is likely to fail over time under its actual material, climate, drainage, repair, and exposure conditions. The difference is timing: the action changes when a risky roof is expected to fail in months instead of years.
Does RAKE ML replace physical inspections?
No. It prioritizes inspections, improves underwriting triage, and gives asset teams a defensible view before they send people into the field. The highest-risk assets still deserve direct investigation.
How do insurers use RAKE ML?
Underwriting teams can use RAKE ML to screen submissions, identify roofs that need inspection, support premium load decisions, plan non-renewal timing, and prepare claims resources where failure risk is concentrated.
How do lenders and servicers use it?
Credit teams can monitor collateral condition, compare asset health against reserve expectations, and flag deterioration before a borrower issue becomes a workout.
What happens in a portfolio assessment?
We review the portfolio, available data, buyer workflow, decision threshold, and the operational action the report needs to support. The first output is scoped so your team can judge whether the signal is useful before broader deployment.
Start here
Bring physical truth into the next underwriting decision.
Tell us whether you are underwriting, lending against, insuring, brokering, or operating the portfolio. We will respond with the cleanest first step.
- Portfolio and single-asset scoping
- Underwriting and credit workflow review
- Pilot design for insurers, lenders, owners, and brokers
We've got it. Thank you.
Our team will follow up within one business day with a practical next step.
Want to get in touch faster?
Book a 30-Minute Call