Specialty Chemicals Data Guide: 20 Datasets & Sources (2026)

Specialty chemicals data clusters into five families: commercial price intelligence, regulatory inventories, official production statistics, cheminformatics reference sets and sustainability impact data. As of August 2026, Datadory catalogs 20 datasets for the industry — 12 primary sources and 8 related records drawn from neighboring industries such as commodity chemicals, diversified chemicals and biotechnology. Eleven of the 12 primary datasets are free outright, and 3 expose an official API.

Official statistics handle market sizing. The 2022 US Economic Census puts NAICS 325 chemical manufacturing at 14,961 establishments and $901.0 billion in value of shipments. UK PRODCOM ships final estimates as a single XLSX whose Table 1 reaches back to 2006, India's data.gov.in aggregates 68 chemical and petrochemical visualizations, and FAOSTAT tracks pesticide trade for 233 countries and territories back to 1961.

Cheminformatics depth is unusual for an industry vertical: Zenodo's InChIKey-Deduplicated ClassyFire/ChemOnt Label Collection carries class labels for 73,356,229 compounds as roughly 4.8 GB of Parquet, Kaggle adds 426,850 molecules with RDKit descriptors for lipophilicity prediction, and OECD eChemPortal federates about 800,000 substance records from 34 participating government databases.

Twelve sources do the heavy lifting. Three of them — ECHA CHEM, the EPA TSCA Inventory and the NAICS 325 Economic Census package — score 9 on Datadory's 10-point quality rubric, and the twelve together average 7.4 against the catalog-wide mean of 7.81.

Government and regulatory inventories

EPA TSCA Chemical Substance Inventory (quality score 9) is the official list of substances manufactured, imported or processed in US commerce under TSCA. The July 2026 snapshot carries about 70,800 non-confidential listings — 36,522 active and 34,252 inactive — alongside roughly 16,000 confidential-entry records, downloadable as a 2.5 MB CSV zip or MS Access file. EPA has published the inventory since 1979 and refreshes it about every six months; the public version is explicitly not authoritative for compliance, since only the confidential Master Inventory File is definitive.

OECD eChemPortal (quality score 7) runs one search simultaneously across 34 national and regional government databases holding about 800,000 substance records, surfacing physicochemical properties, hazard data, environmental fate, exposure information and GHS classifications through a free API that returns up to 20 results per page.

Official production statistics

US Economic Census — Chemical Manufacturing (NAICS 325) (quality score 9) is the market-sizing backbone: 14,961 establishments and $901.0 billion in value of shipments for reference year 2022, with employment broken out to six-digit industries and geography down to county level. Roughly 3,600 NAICS 325 rows ship in the sector31 basic package as a ~4.3 MB zipped CSV, released every five years under US public-domain terms.

data.gov.uk PRODCOM Final Results (quality score 7) packs UK manufacturers' sales value and volume across 25 divisions — including Division 20 chemicals — into one ~1.4 MB XLSX of thousands of eight-digit product codes. It carries 2025 estimates, revised 2024 figures and finals for 2015-2023, holds National Statistics designation, and is cataloged by Datadory.0.

FAOSTAT Pesticides Trade & Use (quality score 8) delivers annual pesticide import and export values and quantities for 233 countries and territories from 1961 to 2023 — values only before 1990, quantities from 1990 onward — pulled as CSV or JSON through a BigQuery-backed API under commercial delivery terms 4.0.

Commercial price intelligence and sourcing

ChemAnalyst - Chemical Price Database (quality score 6) tracks daily price movements, forecasts, supply-demand balances and plant-level data for roughly 1,100 chemical and commodity products across more than 40 countries. Its 1,098 sitemap-listed public pricing pages display the quarter ending June 2026 for free; daily-granularity series and forecasts out to ten years sit behind subscription.

ChemicalBook - Chemical Database & Supplier Directory (quality score 8) indexes records by CAS number, name and structure, pairing properties, synthesis routes, uses and MSDS links with a global supplier directory weighted toward Chinese manufacturers. Its sitemap index lists 1,499 catalog URLs, the site reports 136,861 registered members, and robots.txt permits product and CAS paths.

Cheminformatics and sustainability reference sets

InChIKey-Deduplicated ClassyFire/ChemOnt Label Collection (quality score 7) carries ClassyFire/ChemOnt class labels for 73,356,229 unique compounds drawn from PubChem, ZINC20 and Enamine REAL, distributed as ~4.8 GB of Parquet train/validation/test splits under commercial delivery terms 4.0 (version v4, published June 2, 2026).

Kaggle — 426k PubChem Molecules for LogP Prediction (quality score 6) supplies 426,850 bioactive molecule rows with SMILES/InChI identifiers, eight RDKit 2D descriptors and an XLogP3 target — curated as the intersection of PubChem, ChEMBL and eMolecules — in ~33 MB zipped under commercial delivery terms-NC-SA 4.0 terms (version 1 uploaded November 20, 2025).

Each major specialty chemicals workflow maps to specific named sources rather than generic categories:

1. Shortlist suppliers per compound. ChemicalBook pairs CAS-indexed properties, synthesis and uses with a global supplier directory and contacts — 1,499 sitemap-listed catalog URLs, crawlable because robots.txt permits product and CAS paths.

1. Screen US commerce status and compliance exposure. The EPA TSCA Inventory's ~70,800 non-confidential listings split into 36,522 active and 34,252 inactive substances, refreshed about every six months as a 2.5 MB CSV zip.

1. Size and rank local US chemical markets. The 2022 Economic Census NAICS 325 package gives establishment counts (14,961), shipment values ($901.0 billion) and employment detail down to county level across roughly 3,600 rows.

1. Quantify UK chemical output. PRODCOM's Division 20 sales value and volume series arrive as National Statistics in a single ~1.4 MB XLSX, with Table 1 extending back to 2006.

1. Track global agrochemical trade flows. FAOSTAT covers pesticide import and export values and quantities for 233 countries and territories, 1961-2023, country by country and product by product.

1. Train cheminformatics models. The Zenodo ClassyFire/ChemOnt collection labels 73,356,229 compounds in ~4.8 GB of Parquet splits; Kaggle's 426,850-row set validates lipophilicity predictions against an XLogP3 target.

1. Estimate product carbon footprints. Dryad's commercial delivery terms life-cycle impact distributions span GWP, acidification, eutrophication, ozone depletion and POCP by chemical taxonomy, derived from Federal LCA Commons processes.

1. Compile multi-jurisdiction hazard dossiers. OECD eChemPortal fans one query out simultaneously across 34 government databases covering about 800,000 substance records, returning up to 20 results per page.

Four kinds of teams drive most specialty chemicals data searches, and each starts from a different source:

Data scientists arrive hunting training data. Their queries look like "classyfire chemont labels dataset parquet" or "pubchem logp dataset csv," and the answers live in the Zenodo collection's 73,356,229-compound Parquet splits and Kaggle's 426,850-molecule descriptor table. Of the 1,744 datasets Datadory catalogs, 666 reach top-tier relevance for modeling work, and 620 of those are completely free. Our data scientists page collects these modeling workflows.

Market researchers need market-sizing anchors, not transactions. They search phrases like "naics 325 chemical manufacturing statistics" and "uk prodcom division 20 data," landing on the Economic Census's 14,961-establishment NAICS 325 cut and PRODCOM's Division 20 series. Across the catalog, 786 datasets rate as strongly relevant for this persona, 84.6% of them free, and 75.1% score 8-plus on quality. The market researchers page maps the sizing workflow.

A practical specialty chemicals data stack builds in four layers, cheapest and most durable first:

Every figure below comes straight from the cataloged source records, as of August 2026. The industry-specific rows come from the 12 primary specialty chemicals datasets; the catalog-wide rows cover all 1,744 datasets Datadory catalogs.

Three numbers carry the most planning weight. First, the free-share gap: this slice runs 91.7% free against 82.4% catalog-wide, so budget conversations start late here. Second, the API gap: 25% of primary datasets expose APIs against 32.9% catalog-wide, so expect batch files rather than live feeds. Third, the scale spread: 73,356,229 labeled compounds on one end and $901.0 billion in shipments on the other mean the narrowest and broadest questions in the industry are both answerable from free sources.

Continue into the connected pages: