US EPA Facility Registry & Commercial Environmental Risk Matcher

Registry matching and review prioritization using public EPA facility records.

Find U.S. facilities in the Environmental Protection Agency's Facility Registry Service (FRS), organize their recorded NAICS codes and program associations, and optionally compare a facility name and address with the registry.

“Risk matcher” describes record matching and the prioritization of records for further review. It does not mean an environmental-risk score, a violation detector, a compliance certification, or a determination of liability. An FRS record or program association alone does not establish a violation.

What you receive

• One Dataset item per unique FRS registry ID, with facility name, address, industry codes, program associations and available coordinates.
• A direct EPA source URL and observation timestamp for each record.
• Explainable match status, heuristic confidence and field-level evidence when you supply a name or address.
• A free run summary stating why processing stopped and whether the queried source was exhausted.
• JSON and CSV exports through Apify Dataset.

This release does not retrieve separate compliance findings. compliance_status and compliance_status_url remain null. Missing coordinates remain null and are not converted to 0,0.

Quick start

The small default query is Colorado ZIP 80002, at most five unique facilities and a $0.03 facility-charge cap.

{
  "state": "CO",
  "zip": "80002",
  "maxResults": 5,
  "maxCostUsd": 0.03
}

To compare a known facility against EPA records:

{
  "state": "VA",
  "zip": "23234",
  "facilityName": "J W FERGUSSON & SONS, LLC",
  "streetAddress": "4107 CASTLEWOOD ROAD",
  "maxResults": 1,
  "maxCostUsd": 0.006
}

Inputs

Field	Meaning
state	Two-letter U.S. state or territory abbreviation.
zip	Five-digit ZIP or ZIP+4, supplied as text. A five-digit query includes ZIP+4 records.
naicsCode	Two to six digits. A prefix matches any associated NAICS code beginning with those digits.
epaProgram	An FRS program acronym. Program associations are registry facts, not noncompliance findings.
facilityName	Optional identity comparison. Matching results require review as described below.
streetAddress	Optional address comparison, including building and unit number when available.
maxResults	1–1,000 unique delivered facilities; default 5.
maxCostUsd	Maximum facility charges in USD; default 0.03. The lower Apify run spending limit also applies.

Multiple supplied filters are combined. Industry or program filters identify candidate facilities, while enrichment preserves the associated values retrieved for each facility. State and ZIP filter the public master query. NAICS and program filters are checked after complete program/industry enrichment, independently at the facility level. Narrow geographic filters reduce requests and processing time. If all filters are omitted, the safe default geography is used.

Output fields

Field	Interpretation
registry_id	EPA's 12-digit registry identifier, preserved as a string.
facility_name	Recorded primary facility name.
street_address, city, state, zip	Recorded facility location. ZIP remains text to preserve leading zeros.
naics_codes	Deduplicated recorded NAICS codes; an empty array means none were returned.
epa_programs	Distinct FRS program acronyms returned for this facility.
epa_program_details	Program acronyms with their source program identifiers.
latitude, longitude	Available EPA coordinates, or null when absent/invalid.
source_url	Public EPA source query for the facility.
observed_at	UTC retrieval timestamp; not a certification that the underlying record was updated on that date.
match_status	not_requested, exact_match, probable_match or possible_match for delivered records.
match_confidence	0–1 heuristic similarity score, or null; not a statistically calibrated probability.
match_evidence	Query/source values, normalization comparisons, agreements and reasons for review.
compliance_status, compliance_status_url	Null in this release; compliance is not assessed.

JSON preserves arrays and nested evidence. CSV uses Apify's standard serialization of those values; it does not create multiple billed records for a facility's multiple programs.

How matching works

Name comparisons normalize punctuation, spacing, “&” and common legal suffixes. Address comparisons normalize common street abbreviations while retaining street and unit numbers. A conflicting known street or unit number is not accepted as a match.

An exact_match requires both an exactly normalized name and address. Strong address agreement with a similar name can produce a probable_match. A name-only or address-only comparison produces at most a possible_match; an address may be shared by several facilities.

When matching is requested, candidates classified as not_matched are omitted from the billed Dataset. If the scoped search finishes without a match, the free SUMMARY record reports unmatched. This means no accepted match was found in the queried scope; it does not prove the facility is absent from every source.

Pricing and spending limits

$0.006 per unique delivered facility — $6 per 1,000 facilities. Failed requests, duplicate registry IDs and summary/diagnostic records are not additional facility events. The release uses one per-Dataset-item billing event, with no separate startup event or additional platform-usage charge for ordinary runs.

The Actor checks maxResults, the input cost cap and the platform's run spending limit before output. It uses whole-record capacity: a $0.013 cap permits at most two $0.006 records. The run summary's charge estimate is the record-price calculation; Apify's billing/run details are the authoritative charge record.

Reliability and continuation

EPA requests use 1-based inclusive pages sorted by registry ID, a 25-second request timeout, a maximum of two retries, at least 500 ms between requests, and a 250-request run ceiling. A run has a four-minute processing allowance and a 10,000-candidate safeguard. HTTP 429 responses respect Retry-After; a delay beyond the run allowance stops processing rather than retrying early. Persistent errors fail visibly while retaining progress. If EPA returns a row outside the requested state or ZIP5 scope, the Actor fails before enrichment instead of silently expanding the search.

Progress includes the pending page and candidate position. On resurrection, the Actor reads existing Dataset registry IDs before continuing, preventing a record accepted before an interrupted response from being blindly written again. Durable input signatures prevent resuming with different filters into the same Dataset. Start a new run to change filters.

A complete unlimited national export is outside this Actor's run scope. The Actor stops at the result/budget limit; source failures or a runtime/candidate safeguard report incomplete processing. The summary separately states whether the queried source was exhausted.

Data sources and terms

• EPA Envirofacts Data Service API
• EPA FRS public API
• EPA FRS dataset catalog and license
• EPA Standard Open Data License
• EPA's distinction between public and intergovernmental FRS APIs

The FRS catalog marks the dataset public and links EPA's public-domain data license. Use is subject to the applicable metadata and EPA's accuracy/utility limitations. This Actor uses public query data and does not submit data to EPA or access NAAS-protected intergovernmental records.

FRS integrates records from multiple programs and can contain incomplete, inconsistent, historical or periodically refreshed information. A program association may refer to a historical interest. Source absence is not evidence of compliance. This independent product is not endorsed by the U.S. EPA.

Verified examples

Live EPA validation on 19 October 2026 covered Colorado and Virginia. The five-record Colorado default completed in 14.6 seconds on the development server. Virginia registry 110000869588 matched the normalized facility name and street address with an exact_match and a 0.99 heuristic score. Colorado registry 110002993226 was retained with both coordinates null. A nonexistent supplied identity was compared with all 404 records in Colorado ZIP 80002 and produced an empty Dataset and an unmatched summary. The Apify default run also completed in about 5.5 seconds. A forced cloud timeout after 23 delivered records was resumed to 40 unique records and exactly 40 billed Dataset events, without duplicate output. These timings are observations, not service guarantees.

Version and support

Version: 1.0.0. Project code: NEWAPI10. Report issues with the run ID, nonsecret input and an example registry ID. Never send access tokens in issue reports.