Parcel-level buildability scoring layer over US public records: every residential parcel is screened for flood, slope, soil, access and zoning constraints, with each finding carrying its source and the date it was read, and unresolvable items explicitly flagged 'Needs verification' rather than defaulted to ready. Evidence is assembled from 7 federal agencies plus 5 commercial and local sources (FEMA, USGS, EPA, Census and others per the published source catalog), then fused into a Buildability Score whose per-factor sub-scores are stored, and backtested against a labelled built-vs-never-built parcel dataset.
Equities in US. Related tickers LEN, DHI, PHM, KBH, LGIH, CBRE, AVB.
Tabular + geospatial (parcel-level factor tables and scores, delivered per address/jurisdiction and embeddable on a city's own site) + Text (cited source records and open-question lists naming what could not be read)
Per-parcel screening across every residential parcel in each supported city (Austin is the reference market, with the published backtest cohort alone being 76 built + 369 undeveloped parcels), each parcel carrying 5 constraint factor sub-scores, source citations and read-dates, drawn from a 12-source catalog (7 federal agencies + 5 commercial/local).
The rarest asset here is a hand-cleaned supervised label set the company published: 76 completed new-construction permits paired against 369 undeveloped parcels of the same zoning class and minimum lot size in Austin TX, i.e. built vs never-built ground truth with neighbour-matched pairs, plus stored per-factor sub-scores for every parcel in the cohort and three successive score versions (v1 AUC 0.741, v2 0.706, v3 0.711 with confidence intervals) retained even where the restructure did not improve discrimination. Free labels on the main layer: per-parcel constraint flags (flood / slope / soil / access / zoning) each with source and read-date, a four-state verification status, and an explicit 'open questions' list of unanswerable fields.
Buildability™ (Buildability AI, buildability.us) — Parcel-level buildability scoring layer over US public records: every residential parcel is screened for flood, slope, soil, access and zoning constraints, with each finding carrying its source and the date it was read, and unresolvable items explicitly flagged 'Needs verification' rather than defaulted to ready. Evidence is assembled from 7 federal agencies plus 5 commercial and local sources (FEMA, USGS, EPA, Census and others per the published source catalog), then fused into a Buildability Score whose per-factor sub-scores are stored, and…
Not yet on file: Sample · Data dictionary · Data frequency · Reporting lag · Point-in-time · License