── ── Open data
US Business Registry Open Data
Five US states — New York, Colorado, Pennsylvania, Oregon, and Connecticut — publish their entire business registries as open data with explicit public-domain or commercial-OK licenses: roughly 12.4 million entities via documented Socrata APIs. This skill ships the verified dataset registry, a dependency-free Python fetcher, measured rate limits, and the gotchas that corrupt naive pulls.
Run US Business Registry Open Data on a real problem
Bring something you're actually deciding — free, in the browser.
How it works
Pick states from the verified dataset registry and gate on license: a dataset with no explicit public-domain or commercial-OK tag stays off — a portal listing is not a license.
Pull with plain $limit/$offset pagination ordered by :id at ~500 rows/sec anonymously. Keyset pagination via $select=:*,* and the CSV export endpoint both measured dramatically slower — the bottleneck is per-page transfer, not offset depth.
Normalize onto a unified schema, dedup by entity ID (Oregon is row-per-name, Pennsylvania row-per-officer — naive counts overcount 2–3×), and resume interrupted pulls by line count in append mode.
When to use it
- Building lead lists or TAM sizing from state registration data
- Formation-trend, entity-type, or business-survival analysis by cohort
- Entity matching and enrichment against official registry records
- Deciding whether to pay OpenCorporates or a data vendor — check the free floor first
When not to use it
When you need business license data (a different registry), SEC filings, non-US registries, or full national coverage including Delaware, California, and Texas — no free path exists for those; budget for a vendor.
Worked example
Pulling 12.4M registrations across five states
A June 2026 pull verified all five datasets live with count(*) queries, caught Iowa's platform migration (every documented endpoint 404), left unlicensed Hawaii disabled, and measured the pagination options: plain offset ~500 rows/sec, keyset a 200-second timeout, CSV export 12× slower. Deduping Pennsylvania's officer-level rows turned an inflated row count into 2.31M honest entities.
Why it matters for founders
Business registries are the skeleton of the US SMB economy — formation trends by county, registered-agent market share, survival curves by cohort. deciqAI pulled this data for company matching in its analysis agents; the same free datasets can seed your lead lists and market sizing before you pay a data vendor a dollar.
Install this skill (free, MIT)
npx skills add deciqAI/knowledge-skillsUseful? Star the repo — stars help other builders find it.
FAQ
Is this data legal to use commercially?
Yes, for the five enabled states — each dataset carries an explicit public-domain or commercial-OK license verified on its portal page. Datasets without a clear license tag (like Hawaii's) stay disabled until terms are confirmed.
Why isn't Delaware included?
Delaware — where a huge share of US companies incorporate — offers no bulk product and no API at any price. Selling registry data is part of the state's business model. California and Texas charge for bulk access.
How long does a full pull take?
At the measured anonymous rate of ~500 rows/sec, all ~13M rows take 7–8 hours. Registering a free Socrata app token and sending it as X-App-Token raises the rate-limit tier.
Do I need any dependencies?
No — the bundled fetcher is ~300 lines of Python stdlib. Each state is one entry in a config dict; adding another Socrata state is one more entry.
Related mental models
Strip a problem to statements that can't be reduced further — physical law, definition, cited fact — then rebuild the answer using only those.
When several explanations all fit the evidence, prefer the one that assumes the least. A selection heuristic for what to bet on first.
First-level thinking asks “what happens?” Second-order asks “…and then what?” — catching the effect that reverses the obvious one once everyone responds.
Instead of “how do I win?”, ask “how could this fail catastrophically?” — then design the plan around eliminating the failure paths that matter most.