Why a land firm built its own parcel search instead of licensing one.

County GIS systems publish parcels, zoning, flood zones, soils, easements, and ownership. The problem is that every county publishes them separately, with different field names, different layers, and different gaps. Evaluating a site meant opening one county's map viewer, then another, cross-referencing records, and measuring boundaries by hand. The firm had already looked at a commercial land-data product. It wanted something it owned and could tune to its own criteria.

We started with one county and the firm's literal brief: press a button, get a list of parcels between 5 and 100 acres. Seven days after the scoping call we demoed a working map.

The firm's operations lead sent back a page of edits: take size and zoning out of the score because they are filters, add a utilities dimension, and split environmental into streams, flood-prone soils, and flood hazard. Then the founder sent the weights. We built both of his profiles exactly as written. Major subdivisions weight environmental constraints heaviest. Minor subdivisions weight parcel shape heaviest. Where a county does not publish a data layer, the score redistributes that weight instead of guessing in the parcel's favor.

The firm asked for eight counties. One sits behind a firewall, so we shipped seven. Another took its parcel service offline mid-project, and we moved it to the state's parcel feed, which carries less detail. Four counties do not publish road centerlines, so road frontage there is an estimate, marked with "≈" on screen. The firm also caught a measurement bug: one parcel's longest boundary read 334 feet when the county's own viewer showed about 1,736. We were measuring a single segment instead of the full straight run. We fixed it, and the firm got a new filter on that measurement out of it.