Where Recall Data Comes From — and the Limits of Public Records

If a site tells you which products are dangerous, the first question worth asking is where it gets its information. Ours comes from two public sources, both run by US authorities, and both free to query.

Source one: the CPSC recall feed

The US Consumer Product Safety Commission publishes recall notices for consumer products, and exposes them as machine-readable records. Each record carries the notice title, the full description, the products involved, the hazards and injuries associated with the recall, the publication date and a link to the official notice.

That richness is why consumer product alerts on this site are the most detailed: there is genuinely more to say. It is also why these alerts can carry a photograph — the notice includes images of the recalled item, and the agency’s terms allow copying recall notices, photographs included, with credit.

Source two: openFDA enforcement records

The FDA publishes recall enforcement reports as structured data on openFDA. These cover food, drugs, medical devices, cosmetics and tobacco products. The records are generated from firm submissions, and their vocabulary is administrative: firm name, product description, catalogue or model numbers, lot codes, distribution area, classification, reason for recall.

The limits are worth stating plainly. These records do not include product photographs. They describe firm-initiated recalls, including the long cascades where one supplier’s defect recalls dozens of downstream products. And the descriptive text is often a catalogue number list rather than a consumer-facing description.

What neither source does

They do not know who bought the product. No register links a household to a unit. That is the delivery gap that alerts exist to narrow.
They do not always cover a category you expect. Some products are covered by different regimes entirely: vehicles by NHTSA, workplace equipment by OSHA, most consumer goods only if they fall under a specific statute.
They are not instant. A recall published by an agency can lag the company’s internal decision, and enforcement records in particular appear in batches.
They are not consumer-friendly documentation. That is the gap we try to close with summaries and localised pages, without touching the facts.

The honesty rules we apply to ourselves

We never invent a field. If the source record has no image, the alert has no image. If it does not state a severity, we do not manufacture one from the headline.
We quote rather than paraphrase for the hazard statement. A rewording could turn “can cause internal burns” into “might be hot”.
We link the source on every alert. For medical devices, medicines and food, the official record is the one to act on — our card is a signpost.
We date everything. A recall record without its publication date is useless: notices get extended, remedies change and records are corrected.

Why a third stream is quiet

We also had a complaints stream in the architecture. It is currently failing: the CFPB endpoint refuses our requests, so those records simply do not appear. Rather than fill the gap with something plausible, we leave it empty and show the two sources that work.

You can check any of this directly. Every alert links to its publication, and the raw list is browsable — start with the current alerts, or subscribe to the categories you care about.

Sources: CPSC recalls · openFDA · openFDA data licensing