Skyline

Executive summary

Knight Commercial manages a property book spread across seventeen states. When a storm lands, the commercially valuable window is short and the question is narrow: which of our properties were just inside that polygon?

Answering it by hand meant watching the weather, guessing at geography, and reconciling against a spreadsheet — slow enough that the window usually closed first. Digital Boutique built Skyline: a hazard-intelligence layer that watches the NOAA/NWS feed continuously, resolves each alert polygon against every geocoded property, and produces the outreach list automatically. The whole portfolio sits on one live national map at maps.knightai.space.

The challenge

Storm response in commercial property is a race with a decaying prize. The operator who reaches an owner first, with specifics, wins the work. The one who arrives a day later is competing on price against someone already on site.

Three things made speed structurally impossible. The weather had to be watched by a person, so coverage stopped when they did. Impact was estimated by city or region, which is the wrong geometry — a hail swath is a polygon that cuts across municipal boundaries and misses half of what a city-level match would flag. And the portfolio lived in a spreadsheet, so turning "there was a storm in the DFW corridor" into a list of affected properties meant manual cross-referencing under time pressure, which is exactly when that work is done worst.

The requirement was to remove the person from the detection path entirely, and to make impact a computed geometric fact rather than a judgment call.

Approach

Detection runs continuously. The NOAA/NWS feed is polled every fifteen minutes, so an event is known within minutes of being issued regardless of the hour.

Impact is then resolved geometrically rather than approximated. Alerts arrive as polygons, and every managed and prospect property is geocoded, so the system computes precisely which properties fall inside a given swath. This is the part that changes the answer: a hail event across the Tarrant–Dallas–Collin corridor produces an exact property list, not a regional guess that over-reports in some places and misses entirely in others.

The output is an artifact rather than a dashboard state. A first-notice digest fires on every new event naming the properties in its path, so the work product exists before anyone has opened the map.

The map is where the book is understood rather than where the alerting happens — every property across five portfolios on one national view, with active events overlaid as they land, and a coverage assistant for questions the filters do not answer directly.

On showing this work

The product is access-controlled, and a portfolio map is exactly the kind of asset that cannot be screenshotted for a portfolio site. Skyline publishes its own public preview built on anonymized sample data, and the proof module here follows the same rule: the geometry is schematic, the figures mirror that public preview, and no real portfolio or address appears in it.