FDA Recall Match Delta — Food Watchlists

Run daily or weekly. Pay only for newly seen FDA food-enforcement records that match your products, suppliers, brands, or distribution footprint.

buyer wedge

Built for food distributors, private-label retailers, marketplace compliance teams, and food-safety consultants who need a narrow internal screening feed—not another bulk recall dump.

why it is different

• signal, not retrieval: unmatched FDA records are omitted and free.
• explainable matching: every event lists the field, watch term, and deterministic match method.
• correction-aware: a materially changed source record emits SOURCE_CORRECTION once.
• distribution matching: state abbreviations, state names, and explicit channel phrases are normalized without substring guessing.
• replay-safe: stable event IDs and provider idempotency keys suppress duplicate charges.
• bounded spend: maxEvents and maxTotalChargeUsd cap each run before charging.
• audit evidence: every row includes the official API query, source fingerprint, timestamp, and disclaimer.

source and scope

The Actor queries the official openFDA food enforcement API. No API key, browser, proxy, personal data, or AI inference is required. FDA states that enforcement data is generally updated weekly and is not maintained as an authoritative recall-lifecycle tracker. Accordingly, this Actor is an internal screening aid and emits new or corrected source records; it does not issue public alerts, provide medical/legal advice, or claim a recall was terminated or escalated.

matching

At least one narrow watch term is required:

• keywords: phrase match across product description, recall reason, and code/lot information
• firms: normalized exact or phrase match against recalling firm
• brands: phrase match in product description
• states: exact firm-state or tokenized distribution-area match
• distributionTerms: phrase match in distribution text
• classifications: optional exact filter (Class I, Class II, Class III)

baselineMode=emit_matches returns matching records in the initial lookback window. Use silent to establish state without emitting or charging initial matches. Reuse the same watchlistId, settings, and stateStoreName for scheduled monitoring.

output

Paid dataset rows are either:

• NEW_MATCH — $0.20
• SOURCE_CORRECTION — $0.05

Unmatched records, unchanged replays, silent baselines, malformed input, and failed source fetches are not paid events. A deterministic run report is stored as REPORT with source window, counts, receipts, state size, and claim scope.

operational note

If openFDA returns an error or an unexpected response shape, the run fails before state is advanced or events are charged. Event settlement is fail-closed: an ambiguous prior settlement refuses another provider charge.

quick start

Start with a silent baseline so historical matches do not create paid events, then schedule the identical input daily or weekly.

{
  "watchlistId": "private-label-snacks",
  "brands": ["Example Brand"],
  "firms": ["Example Foods LLC"],
  "keywords": ["undeclared peanut"],
  "states": ["WA", "OR"],
  "classifications": ["Class I"],
  "lookbackDays": 14,
  "baselineMode": "silent",
  "maxEvents": 10,
  "maxTotalChargeUsd": 2,
  "stateStoreName": "private-label-snacks-fda"
}

After the baseline, keep watchlistId, watch terms, and stateStoreName stable. Route returned NEW_MATCH and SOURCE_CORRECTION rows to a compliance review queue. Confirm every action against the linked FDA source; this feed is deliberately not an emergency notification or recall-status system.

Useful workflow: Apify schedule → this actor → dataset webhook → Slack, email, ticket, or supplier-review queue. The event includes the matched field and term, so reviewers can see why it appeared without an AI explanation layer.

Other focused monitors: local website-gap changes, Florida new entities, and Seattle pre-permit events.