FDA Import Refusal & Supplier Risk Signals

Turn the FDA's official OASIS Import Refusal Report CSV archives into structured historical shipment-line signals for supplier review, due diligence, Clay and n8n. The Actor retrieves FDA ZIP archives directly over HTTPS and exports one unique refusal line per Dataset row in JSON or CSV.

Source and coverage

• Official report: https://www.accessdata.fda.gov/scripts/importrefusals/index.cfm
• Official CSV ZIP: https://www.accessdata.fda.gov/scripts/importrefusals/downloads/Import_Refusal_2024-present.zip
• Official field dictionary: use "Structure of the Import Refusal Data Files" on the report page. The CSV contains MFG_FIRM_FEI_NUM, LGL_NAME, address, ISO_CNTRY_CODE, PRODUCT_CODE, REFUSAL_DATE, DISTRICT, ENTRY_NUM, RFRNC_DOC_ID, LINE_NUM, LINE_SFX_ID, sample flags, REFUSAL_CHARGES and PRDCT_CODE_DESC_TEXT. ACT_SECTION_CHARGES.csv maps ASC_ID to charge code, statement and section.
• FDA posts archive sets for 2002–2006, 2007–2010, 2011–2013, 2014–2018, 2019–2023 and 2024–present. An archive is downloaded only when its years intersect the requested interval. The current 2024–present ZIP inspected on 2026-11-03 contains REFUSAL_ENTRY_2024-Sep2026.csv, with refusal dates through 2026-09-24. Publication is monthly and can lag the current calendar date. No live status is inferred.
• FDA says the report concerns products whose admission was refused in whole or part. LGL_NAME is the declared manufacturer, the establishment declared responsible for the product; it does not prove the actual manufacturer, shipper, importer or supplier relationship. The Actor sets firm_role=declared_manufacturer. Every row is a particular entry line, not a blanket ban, current sanction or Import Alert.

Inputs

Field	Behavior
refusalDateFrom / refusalDateTo	Inclusive YYYY-MM-DD; default last 365 days through UTC today; max ten-year interval; earliest CSV archive 2002.
country	Exact two-letter ISO manufacturer country code (e.g. CN), not a free-form country name.
firmName	Literal case-insensitive substring in LGL_NAME.
productKeyword	Literal case-insensitive substring in FDA product description.
industryCode	Exact first two digits of PRODUCT_CODE, e.g. 16.
maxResults	At most 10,000 unique entry lines; default 10.
maxCostUsd	Additional local event budget; default $0.08, max $80; Apify's per-run limit also applies.

All filters combine with AND. No unsupported openFDA import-refusal endpoint is used. A zero-match run produces an empty Dataset and zero result events. The default reads the latest archive and stops after ten matching rows, normally in under five minutes.

Dataset

Each row includes refusal_id, firm_name, firm_role, firm_address, country, product_description, product_code, industry_code, refusal_date, refusal_charges, source_url and fetched_at. The entry, reference document, line and suffix remain separate, and source_record retains the original FDA CSV fields. refusal_id joins ENTRY_NUM, RFRNC_DOC_ID, LINE_NUM, LINE_SFX_ID and REFUSAL_DATE; it never merges different shipment lines based on company name. The source_url is the downloadable archive URL, since FDA does not provide a stable per-CSV-row URL.

refusal_charges is an array of objects with asc_id, code, statement and section, looked up by ASC_ID. Repeated, comma-separated reasons remain in one row. If a newly published charge ID is absent from its dictionary, its identifier is retained with null translation fields. No additional result events are generated for reasons. The source may change historical records or corrected descriptions; use source_url, fetched_at and refusal_date to interpret freshness.

Pricing and limits

Pay per event: $0.008 per successfully delivered unique refusal line ($8 per 1,000). Event name: unique-refusal. There is no start, reason, retry, download or duplicate charge. The SDK pushes and charges one row in one operation. Existing Dataset IDs are checked after a run restart to avoid repeat events. A new run with the same filters may charge again; deduplication is per run. The Actor stops at the smaller of maxResults, maxCostUsd and Apify's run charge limit. Standard-run compute is included in the event price; platform post-run Dataset access/storage may have separate standard fees.

Downloads are bounded to 25 MiB compressed and 100 MiB per CSV member, use HTTPS, at most three attempts on transient errors, a 35-second request timeout and 210-second overall fetch/scan deadline. Archive ranges act as year-based partitions; filters are applied locally while streaming the CSV, so high-volume broad history queries may require narrower dates. An upstream schema change or failure terminates the run rather than silently returning incomplete successful data. SUMMARY records source archives, counts, limits and completion status. Results already delivered before a failure remain accessible and billable.

Example and integration

{"country":"CN","refusalDateFrom":"2026-09-01","refusalDateTo":"2026-09-30","maxResults":5,"maxCostUsd":0.04}

Call the Actor with JSON and read Output's default Dataset as JSON or CSV. Use refusal_id as a Clay/n8n upsert key. JSON preserves nested charge arrays; CSV serializes nested cells and may need parsing downstream.

Use and attribution

FDA's website policies permit reuse of its public-domain website material unless otherwise noted and recommend linking back and recording the copy date: https://www.fda.gov/about-fda/about-website/website-policies . The Actor provides source links and fetch timestamps, charges for processing, and does not claim FDA endorsement. Verify original FDA records and current regulatory status independently before making business or legal decisions.