makenomistakesllc/us-building-permits-16-cities

Dataset Card: US Building Permits — 16 Cities

43,119 building/construction permits issued across 16 US cities over the 30 days ending 2026-09-22, normalized from each city's official open-data API (Socrata / ArcGIS / Carto). See README_permits.md for the full data card (provenance, collection method, known gaps, license) and SCHEMA_permits.md for the column dictionary.

Fresh, scheduled data (including an incremental "new permits only" mode): the source Apify Actor at https://apify.com/makenomistakes/us-building-permits-scraper — https://github.com/makenomistakesllc/apify-actors

US Building Permits — 16 Cities, 30-Day Sample (Sep 2026)

What this is

43,119 building and construction permits issued across 16 major US cities over the 30 days ending 2026-09-22, normalized onto one flat schema: permit number, permit type, work description, status, issued/applied dates, job-site address and coordinates, contractor name and (where published) phone, declared valuation and a link back to the source record. Every row traces back to the issuing city's own public, official open-data API — there is no scraping of city websites, no login, and nothing here is inferred or corrected by hand.

Why it's useful. A building permit is one of the earliest public signals that money is about to move: a contractor lead before the job is finished, a demand signal for building materials before the order is placed, and a real-estate investment signal months before it shows up in sales comps. Sixteen cities on one schema means you can compare permit velocity, valuation and contractor activity across metros without hand-building 16 separate scrapers.

Provenance

Every record comes straight from the city's own open-data platform:

Collected 2026-09-22 via the us-building-permits-scraper Apify Actor, which queries each city's documented SoQL / ArcGIS where / Carto SQL endpoint directly — no HTML parsing, no proxies, no CAPTCHAs.

Collection window and method

issued_date between 2026-08-24 and 2026-09-22 (30 days), non-construction "administrative" records (garage-sale permits, tree affidavits, short-term-rental licenses, home-occupation registrations and similar) excluded at the source query, matching the Actor's default. Seven high-volume cities (New York, Los Angeles, Chicago, Philadelphia, Boston, San Francisco and Washington DC) were pulled through many short, non-overlapping date sub-windows rather than one 30-day call, to stay under a per-run cost ceiling — every window is deduplicated and the full 30-day date range is represented for every city, but the highest-volume cities are a large, representative sample rather than a literally exhaustive count. See SCHEMA_permits.md for the exact row counts and per-column coverage.

Row counts by city

Columns

See SCHEMA_permits.md for the full column dictionary with per-column null rates. Summary: city, permit_number, permit_type, permit_class, work_description, status, issued_date, applied_date, address, zip, lat, lng, contractor_name, contractor_phone, valuation, source_url — 17 columns, 0 missing (every field the source city doesn't publish is null rather than the column being dropped, so the file has no ragged rows).

Known gaps (a property of the source cities, not of the collection)

• —Contractor phone numbers exist for only 2 of 16 cities. Austin (92% of its rows) and Raleigh (79%) publish a phone number; the other 14 cities publish no phone field at all.
• —Austin's coordinates are ~61% complete — about 44% of Austin's permits ship with no latitude/longitude in the source feed at all. Every other city geocodes 95–100% of its rows.
• —Boston, Philadelphia, Washington DC and Louisville publish no application date — only issued (and, variously, expiration/completion) dates exist in those cities' feeds, so applied_date is null on 100% of their rows.
• —Denver publishes no ZIP code column or ZIP-bearing address string, so zip is null on every Denver row; its coordinates come from the layer's point geometry instead and are 100% complete.
• —Washington DC and Philadelphia publish no job valuation. DC's fee field is a permit fee, not a declared job value, and is deliberately not mapped into valuation.
• —`permit_number` is not a unique row key. Several cities publish one row per trade under a shared job number (e.g. New York splits Plumbing/Sprinklers/Standpipe into separate rows for the same job). Use permit_number + permit_type + address + issued_date as the deduplication key if you need one row per trade, or permit_number alone for one row per job. This dataset has zero duplicates on that compound key.

License

This compilation (the normalization, schema and file) is released under CC BY 4.0 — reuse it freely with attribution. The underlying permit records are public government data published by each city under its own open-data terms (several, including San Antonio, explicitly publish under CC-BY); attribute the originating city when you redistribute their records, and check that city's open-data portal for any use-case-specific terms.

Fresh data

This is a point-in-time snapshot (collected 2026-09-22). All 16 city feeds refresh daily. For a live feed — a specific city, a specific trade via keyword filter, a dialable contractor-phone list for Austin/Raleigh, or an incremental "only new since last pull" daily feed — see the Actor this dataset was built from:

• —Actor: us-building-permits-scraper
• —Source / templates (n8n, Make, Postman, MCP): https://github.com/makenomistakesllc/apify-actors

More public-data tools and free datasets from the same team: https://makenomistakesllc.com

Privacy note (added before publication): owner_name was removed from this public release even though the source cities publish it. The Actor still returns it. No owner phone numbers or emails exist in any city feed we use.