Skip to main content

Ground truth tested against
FEMA NFIP claims data.

Every result validated against real loss events — not synthetic benchmarks. All claims sourced from FEMA NFIP Redacted Claims V2.

100%
High-Severity Claim Detection — Every large loss, flagged.
500 / 500 NFIP claims ≥$250K correctly identified · $1.273B captured · Includes all 27 Zone X high-severity claims FEMA classified as minimal risk.
$1.273Btotal captured
Zone X Detection
n = 500
86%
430 / 500 Zone X loss locations flagged — vs 57% FEMA-only baseline across FL, TX, IL.
Before
57%
After
86%
Hurricane Case Studies
n = 700
Harvey 2017 TX88%$88.5M
Ian 2022 FL91%$452M
Helene 2024 FL97%$274M
Milton 2024 FL72%$75.6M
Ike 2008 TX99%$54.9M
Irma 2017 FL81%$62M
Combined 617/70088.1%$1.007B
Storm Peril & SRL
n = 450
Wind 100%
SRL Proximity 80%
Hail 78%
Tornado 38%
SRL breakdown: TX 92.7% · LA 72.0% · FL 74.7%
CAT Events 2015–2024
n = 1,100 · 11 major events
100%
1,100 / 1,100 claims flagged · $1.073B captured
Perfect detection across every major CAT event in the dataset. Every claim, every dollar — identified.
2017
Harvey Irma
2018
Florence Michael
2019
TS Imelda Mid-Summer LA
2021
Ida
2022
Ian
2023
Idalia
2024
Helene Milton
Storm Peril Address Risk — Tier 2
n = 1,400 automated checks
99.9%
1,398 / 1,400 checks passed across wind, hail, tornado & lightning address-level risk fields.
Built entirely on NOAA Storm Events public data — 800 ring-count + 400 address-risk + 100 lightning + 100 cross-consistency checks. The 2 outliers are explained as low-confidence rural edge cases.
Hotspot Differentiation
KDE annual probability
Tornado 4.5x
Tornado AlleyOKC · Moore · Joplin · Wichita
0.36%
Pacific NW / Coastal CAAvg. annual probability
0.08%
Hail 10x
Hail AlleyOKC · Amarillo · Cheyenne · Denver
2.05%
Coastal / Pacific NWAvg. annual probability
0.20%
Methodology Transparency
vs. black-box models
Fully documented · No black box
Address-level annual probabilities are computed via kernel density estimation (KDE) over NOAA Storm Events history, with bandwidth choices published per peril.
P(annual) = 1 − e−λ
Bandwidths: 25km wind & hail · 12km lightning · footprint-based for tornado tracks.
Field-Level Validation
n = 1,400 automated checks
100%
Ring-Count Consistency
800 / 800
99.5%
Address-Risk Sanity
398 / 400
100%
Lightning Field Consistency
100 / 100
100%
Cross-Peril Consistency
100 / 100

See the validation on your coordinates.

Request a demo — we'll run a live report in under two seconds.

Contact Sales →