Hail Prediction

Most weather services detect hail too generally and too slowly, which leaves roofers, insurers, and outdoor operators reacting to damage that is already done. Flash's hail prediction model changes the timeline. Built on four years of convective storm data across the CONUS (2021 through 2024) using a deep learning Conv-GRU architecture, the model forecasts hail size and location up to 55 minutes before the first stone falls, at 1-km horizontal resolution, with updates every five minutes. The product suite includes the 1-Hour Hail Swath Prediction (where, when, and how big), the Hail Size Forecast (predicted hailstone diameter for the swath), Hail Timing down to the minute, Hail Threat Zones rendered as high-confidence polygons over roads and parcels, and a Hail Lookback Tool that lets you search any address for historical hail with a downloadable PDF report. For roofing contractors, the data feeds directly into a CRM or DataForma, triggering automated alerts and inspection reports the moment hail impacts a warrantied roof. No more guessing whether a storm hit, no more arguing with adjusters about coverage, no more clients lost to competitors who got to the property first. The forecast is built for