Bermwise
Operational climate intelligence
Know what breaks before the weather arrives.
Bermwise traces a forecast through your suppliers, routes, equipment, inventory, and orders—then shows the action with the best quantified payoff.
Backtested against real events
Scored honestly, not cherry-picked
- Hurricane Ida, Louisiana 2021
- 100% hit rate
- Winter Storm Uri, Texas 2021
- A miss — published anyway
- Three quiet-control periods, 2021
- 0% false alarms
Mission
Small businesses get the same operational climate risk modeling enterprises pay consultants for—geospatial hazard exposure, published engineering vulnerability curves, and discrete-event simulation, run against their real suppliers, equipment, and orders. Every number in the report traces back to a public source or a cited model, never a guess.
How it works
One address in. A quantified action out.
One address in
Type your business address. Bermwise auto-derives flood zone, active hazards, county, utility territory, and elevation—no questionnaire.
The cascade simulated
A discrete-event simulation propagates the forecast through your actual dependency graph—suppliers, routes, equipment, buffers.
A quantified action out
An optimizer, not a language model, turns the simulation into a recommendation with a computed dollar figure—like moving Thursday's delivery to Wednesday.
- Storm Forecast Issued
- Route Exposure Detected
- Supplier Delivery Delayed
- Revenue At Risk Quantified
Why trust the numbers
The honest evaluation is the marketing.
Every prediction below is scored against a point-in-time snapshot, replayed with a hard leakage guard that raises an error if any query touches data that didn't exist yet. Lower Brier score and MAE are better.
| Event | Brier ↓ | MAE | Lead time | Hit rate |
|---|---|---|---|---|
| December 2022 winter storm — Chicago | 0.062 | 7.0 °F wind chill | 72.5 h | 100% |
| Hurricane Ida — Louisiana, August 2021 | 0.040 | 40.0 mph | 73.9 h | 100% |
| Pacific Northwest heat dome — June 2021 | 0.010 | 4.0 °F | 75.0 h | 100% |
| Winter Storm Uri — Texas, February 2021 | 0.387 | 6.0 °F | 72.5 h | 0% |
Uri remains a miss, not a result to hide. Before the cited grid-reserve model, Uri's disruption probability was 0.15. The reserve-adequacy screen raises it to 0.3779 and improves the Brier score to 0.387—still below the 0.5 hit threshold. The improvement is real but incomplete, and we say so.
Every parameter cited
No vulnerability model reads a magic number. Every constant resolves to a versioned entry in params/ with a source, a URL, a confidence level, and a last-reviewed date. This is what a real provenance card looks like inside the product:
USACE go-consequences default occupancy catalog, HEC-FIA Galveston damage functionsFEMA Hazus Inventory National Database Data Dictionary Appendix A Table A.1 identifies COM8 as Entertainment & Recreation, Restaurants/Barshigh confidence · standardReviewed 2026-08-06
False-alarm rate on quiet periods
0 / 3
Three quiet-control replays—New Orleans, Seattle, and Dallas in 2021, on days when nothing happened—produced zero false alarms. Alert fatigue is a real failure mode; this is how we measure whether we cause it.
Data sources
Public data in. No proprietary black box.
NWS
Active weather alerts, forecasts, and gridpoint conditions
FEMA
National Flood Hazard Layer flood zone mapping
USGS
Stream gauge levels and ground elevation
NOAA
Historical storm events used to backtest and calibrate the models
NIFC
Active wildfire perimeters and incident data
EIA
Grid demand, forecast, and generation data behind outage risk
OpenStreetMap
The road network behind delivery-route exposure
For teams
Bermwise is free while in early access.
We invite workspaces in small batches so live-data capacity stays reliable. Request an invite and we'll be in touch.
Request access