Bulk parcel data for the whole country

Three delivery paths, compared

All three read the same serving set — a Snowflake row, a Parquet row and an API response for one parcel are byte-identical at the same epoch. What differs is the shape you receive, how much of the catalog comes with it, and how it is paid for.

A fourth route exists for Databricks shops: the propraven_silver Delta Sharing share, provisioned directly against your account with no marketplace listing and no platform fee, included with a Team plan or a data license. See Databricks Delta Sharing.

1. Snowflake Marketplace share

The most complete bulk surface. The full PropRaven dataset — including the larger source-side coverage that has not yet been promoted to the public API (building permits, deed transactions, entity resolution, FEMA flood polygons) — lives in the Snowflake Marketplace listing as nine shares: the wide PROPRAVEN_PARCEL_INTELLIGENCE table (191.3M parcels × 620 fields, one row per parcel), the PROPRAVEN_ONTOLOGY knowledge graph, six pre-filtered vertical bundles (Lending, Insurance, Banking, SFR, CRE, GovTech), and the free PROPRAVEN_PREVIEW sample.

It is a share, not a copy: no data movement, no ETL, and you pay per-row consumption through the Snowflake contract you already have. Search “PropRaven” in the Snowflake Marketplace, subscribe to a listing, and run the sample queries in Snowsight. The listing is currently in review; email us for early access. Listing-by-listing detail is on the Marketplace page and in the v1 docs.

2. Parquet on Cloudflare R2

For teams that want files. PropRaven publishes the canonical datasets to one Cloudflare R2 bucket, propraven-data, organised by dataset and immutable version (parcels-silver/v=YYYY-MM-DD/state=NN/part-00000.parquet). Every release ships a self-describing manifest.json with total row count, per-file SHA-256 and partition layout; geometry is WKB in EPSG:4326, and _catalog.json plus _latest.txt point at the current version. Read it with DuckDB, Polars, Spark, pandas or boto3 — no warehouse required, and no egress charge from R2 itself.

Access is minted per session: POST /v1/r2-credentials with your API key returns STS-style credentials scoped read-only to the prefixes you ask for, with a 12-hour TTL. R2 delivery is in beta — Rhode Island parcels (1.77M rows) are live, California (18.29M) is staged, and the national rollout lands on the same publication cadence as the Snowflake share. Quickstarts for DuckDB and Python are on the R2 page.

-- DuckDB: discover what's published, then count a release
SELECT * FROM read_json_auto('s3://propraven-data/_catalog.json');
SELECT count(*)
FROM read_parquet('s3://propraven-data/parcels-silver/v=2026-05-23-ri-test/state=*/*.parquet');

3. CSV export from search

The fastest path to a filtered list. Run a search at /parcels/search — by state, free-text query and property type — and click Export CSV. The tool writes the first 1,000 matching rows with eleven columns: address, city, state, zip, property type, total assessed value, year built, owner name, last sale date, last sale price and data quality score. For larger pulls, GET /api/v1/search/export returns up to 10,000 rows as CSV with fourteen columns (adding parcel ID, state and county FIPS, zoning, lot size in acres, ownership type, latitude and longitude), filtered by bounds, zoning, owner type, absentee flag, sale recency, flood zone and value range. The API export needs a bearer key because it carries owner names.

curl -H "Authorization: Bearer pz_your_key" \
  "https://api.propraven.com/api/v1/search/export?absenteeOnly=true&valueMin=250000" \
  -o absentee.csv

What is included

Every parcel record carries up to 620 columns resolved from county assessor, recorder and permit records: identity (parcel_id, the county-issued APN), a standardized site address with WGS84 latitude / longitude, owner_name as the current owner of record, total_assessed_value for the current tax year, last_sale_date and last_sale_price, year_built, and the FEMA flood_zone designation — plus the structure, land-use, zoning, demographic and hazard enrichments the schema explorer lists field by field with measured fill rates.

• 191.3M parcels · 110.0M mapped — 110.0M distinct locations carry a drawable coordinate
• 620 columns per parcel, 51 states (all 50 plus DC)
• 180M+ building permits · 10M+ title transfers
• 27M+ resolved owners · 143M+ addresses · 293M+ tax records
• In the Snowflake share: entity resolution, FEMA flood polygons, ACS demographics, FEMA risk, climate scoring, CMBS loans and properties

Counts are published at the 2026-08-30 epoch and rounded down. Coverage varies by county because the sources do; check the jurisdiction you care about on the data page before you size a job.

Refresh cadence

The parcel table is rebuilt weekly, and each delivery surface publishes on its own schedule from that rebuild. Snowflake v1 listings refresh monthly (Insurance quarterly, because the FEMA NRI source is annual), with per-row _LOADED_AT and _SOURCE_AS_OF columns exposing actual freshness. On R2, parcels-silver is published on each ontology release (~biweekly), parcels-wide mid-month and permits daily, each under its own v=YYYY-MM-DD prefix with at least the last four versions kept readable. A CSV export reflects whatever the serving epoch is at that moment. Per-source freshness for every joined dataset is on the data page.

Licensing

Whole-state and national tables are sold as a data license: custom pricing, annual term, invoiced. It covers full state or national tables delivered through Snowflake, Databricks or S3-compatible Parquet, refreshed on cadence and unmetered, with an SLA and priority support. The Snowflake Marketplace share is priced differently — per-row consumption metered through your Snowflake contract, with no per-seat license — and the Team plan ($1,000/mo, 1M lookups) adds bulk export to the API. Use of the data is governed by the Terms of Service; the licensor is PropRaven LLC, 2720 South Blvd, #475, Charlotte, NC 28209.

Evaluate before you license

You do not need to commit to see the schema. PROPRAVEN_PREVIEW is a free 10,000-parcel Mecklenburg County, NC sample of every entity with the same column names and confidence model as the paid catalog, so preview queries run unchanged against any paid listing. The Free API plan gives 1,000 lookups a month with no card, and the data page publishes live counts, the full schema and fill rates from production. If the bulk question is really an application question, the Property Data API may be the better fit.

Frequently asked questions

Can I download all U.S. parcel data at once?

Bulk delivery is licensed rather than downloaded from a page. A data license (custom pricing, annual, invoiced) covers full state or national tables delivered through Snowflake, Databricks or Parquet on Cloudflare R2, refreshed on cadence and unmetered. The Snowflake Marketplace share is the self-serve route for teams already in Snowflake.

Is there a free sample?

Yes. PROPRAVEN_PREVIEW is a free 10,000-parcel Mecklenburg County, NC sample of every entity with the same schema as the paid catalog, and the Free API plan gives 1,000 lookups a month with no card.

What format is the bulk data in?

A Snowflake share for warehouse users; Parquet files (snappy, partitioned by state FIPS, WKB geometry) on Cloudflare R2 with a per-release manifest.json carrying row counts and SHA-256 per file; and CSV from the search tool or the search export endpoint.

How many columns and states are included?

620 columns per parcel across 51 states (all 50 plus DC): 191.3M U.S. parcels, 110.0M of them mapped at the 2026-08-30 epoch.