Product rows only.
Send a small sample with product titles, brands, categories, SKUs, model numbers, UPCs, and public product URLs when available.
The workflow is intentionally narrow: redacted product rows in, source-backed recall review queue out. It is not upload-anything storage, not legal certification, and not a place for customer data.
Send a small sample with product titles, brands, categories, SKUs, model numbers, UPCs, and public product URLs when available.
Files containing obvious customer, payment, credential, or private financial data are blocked before recall matching. The formal privacy policy and pilot terms are published at /privacy.
Concierge sample files are treated as temporary inputs for validation, not long-term customer records.
Only product fields go into the test. No customers, orders, payment data, costs, margins, passwords, API keys, or full database exports.
Obvious sensitive fields or structurally useless files are stopped before they enter the CPSC matching step.
Rows are normalized and compared against official CPSC recall records. The output is suppress, manual review, or no visible match.
The report is checked before delivery so obvious formatting issues, bad matches, and unsafe wording can be caught.
Buyer-ready outputs can include HTML, PDF, and CSV report files with official source links and plain-English reasons for each flagged row.
Reports say whether a visible match was found in scanned official sources at the report time. They do not say a product is safe.
The process stores the redacted sample file, normalized product rows used for matching, generated report files, source links, and brief notes needed to decide whether the workflow has real buyer pull.