LESS DATA CLEANING. MORE DOING.
The reference data you need, without the cleanup.
Government source data turned into flat, documented CSVs you can actually use.
One-time purchase · CSV + metadata · Instant download
2017 industry codes → 2022 industry codes
| naics_2017_code | naics_2022_code | naics_2022_title |
|---|---|---|
| 111110 | 111110 | Soybean Farming |
| 111120 | 111120 | Oilseed Farming* |
| 111130 | 111130 | Dry Pea and Bean Farming |
*Title shortened for display. Preview of actual records.
U.S. Census Bureau
+FDA National Drug Code Directory
+NLM RxNorm
THE DATA SHELF / 01
Choose your dataset. Inspect the structure.
Get back to the work that matters.
Bring legacy industry codes into the 2022 standard. A clean starting point for CRM updates, territory mapping, and database joins.
Prepare 2022 industry codes in Python →
What changed from 2017 to 2022? →
Connect FDA drug package codes to therapeutic groups through RxNorm. ATC levels 1–4, together in one file.
Use the NDC → ATC crosswalk in Python →
ACD October release: what changed? →
See where Census areas cross county lines. County names and FIPS, land-area shares, and a largest-land-county flag in one documented file.
2020 geography. Not a current USPS ZIP or address lookup. Shares describe land area, not population.
Add county information to ZCTAs in Python →
Add product names, labelers and dosage forms to package NDCs. Three related source tables and a local enrichment script that preserves your rows and quantities.
FDA source edition 2026-10-09. Source listing does not establish approval or current availability. No ATC classification or historical matching.
Add product names and labelers to NDCs in Python →
Add population, income, education and housing context to county-coded records. A wide CSV for easy joins, plus a detailed table with source definitions and annotations.
2020–2024 survey period · 2024 geography. Includes Puerto Rico and Connecticut’s planning regions. Dollar measures use 2024 dollars.
Add county demographics in Python →
Add approval details and listed patent and exclusivity records to a drug-product research portfolio. Prepared keys, code definitions and documented joins.
Fixed source capture: October 16, 2026. FDA’s precise coverage cutoff is not stated. Application/product identifiers; no NDC linkage.
Join Orange Book products, patents and exclusivities in Python →
Compare employer-business activity across counties and industries. Establishment counts, employment and payroll, with names, size bands and source flags ready to join.
2023 statistics · 2017 NAICS. Covers 50 states and DC. Published county totals stay separate; unavailable counts remain explicit.
Compare county business activity in Python →
Research applications through recorded submissions, status dates and supporting document links. Prepared relationships and readable source definitions.
Fixed source capture: October 16, 2026. Recorded submission statuses, not a complete regulatory event log. Application identifiers; no NDC linkage.
Research FDA application history in Python →
Secure checkout with Stripe. Your ZIP is available after successful payment. Processing improvements for your source edition are free; new source editions may be sold separately. Update policy.
SMALL DETAILS. BIG DIFFERENCE. / 02
Every download comes with the structure and provenance you need to understand what you’re joining.
Each table documents its unique key. Many-to-many mappings stay intact, so you can handle splits and combinations deliberately.
Source code text and NDC hyphens are preserved. Import identifier columns as text to keep spreadsheet software from changing them.
A metadata file records source URLs, the production date, the row count, and a SHA-256 checksum for the CSV.
BEFORE YOU DOWNLOAD / 03
A ZIP containing documented UTF-8 CSV tables, metadata and usage guidance. After confirmed payment, you receive a private download page and emailed access. Use Find my downloads to recover your link later.
Yes. The files are ordinary, flat CSVs. In Excel, use Data → From Text/CSV and set code columns to Text. Check your CRM’s field names and import rules before loading. Some tables have multiple relationships per code. Follow the documented keys and check row counts before summing measures; the FDA reference includes a script that preserves input rows.
No. This October edition classifies 162,167 of 256,748 FDA package NDCs (63.2%), across 632 ATC-4 groups. It uses current prescribable RxNorm concepts and direct ATCPROD product membership. Unmapped codes and exclusions are counted in the included lookup-status table. It includes ATC levels 1–4, not level 5, and is intended for data analysis rather than clinical decisions.
No. It uses 2020 Census ZIP Code Tabulation Areas and 2020 county boundaries, including Connecticut’s former counties. Not every USPS ZIP has a ZCTA, and this file cannot place a street address in a county. It preserves all county overlaps. The shares measure land area, not people or addresses; the primary-county flag selects the largest land overlap.
No. Each price is a one-time payment for the dated snapshot shown. Processing fixes and improvements for your purchased source edition are free. New source editions may be sold separately. Read our update policy.
You’re paying for the extraction, transformation, crosswalk assembly, and packaging. The source agencies provide the underlying data; Actually Clean Data is an independent service and is not affiliated with them.
Chats are processed and stored to provide support. We ask before emailing your conversation to Ben and Hugh. Keep payment details and private download links out of chat. Privacy · Email support
Allow optional Google Analytics to measure visits, sample downloads and checkout clicks? We never send your chat messages, email address or private download link to Google. Cloudflare provides separate cookieless traffic and speed measurements. Privacy details