BL Bestate property intelligence

The brief.

Property data lives in hundreds of separate county systems with inconsistent, frequently broken GIS endpoints. Our own early stack had real defects we had to engineer out: app builds drifting apart, a developer-grade search flow, owner-privacy risk if county detail ever touched our servers — and, most embarrassingly, a public statistics endpoint that quietly stopped telling the truth.

The original build record.

• Problem: county GIS servers return broken JSON, dead pages, and partial results. Answer: a fault-tolerant harvester — JSON repair, pagination fallback, dedup and suppression.
• Problem: public county records can contain owner names, situs and mailing addresses, and parcel coordinates. Answer: the hosted index returns only the fields needed for a user-directed property query, while the buyer's workspace, notes, exports, routes, and enrichment stay under the buyer's control.
• Problem: our published coverage numbers sat frozen for twenty days while the index kept growing underneath them. Root cause: the statistics endpoint read a cache table that a human had to remember to refresh, so a harvest that had already landed was invisible to everyone reading the site. Fix: refreshed from the live index and tied the refresh to the harvest instead of to somebody's memory. A number that requires a person to remember it is a number that will be wrong.
• Problem: three divergent copies of the app drifted apart. Answer: one canonical download plus an in-app updater.
• Problem: the search front door required pasting raw data. Answer: a guided, database-first List Builder with server-side filters.
• Rule we keep: only live production numbers appear publicly — the figures below are fetched from the production API on render, not typed in. Coverage is county-scoped and stated as such; we make no national-coverage claim.