This page explains a reference format for dated business records. It does not offer a downloadable StumbleBees dataset or a live export endpoint.
For the working directory, use the public listings. For the reference specification, read the agent contract.
What an export would need to document
A useful snapshot should identify the record, its public destination, the review date, the evidence behind that date and the method used. The artifact also needs a version, licence and clear description of what its fields mean.
A schema is a contract to validate, not evidence that a particular dataset meets it.
Unknown values
An unknown value should be null or explicitly absent according to the documented schema. It must not silently become zero, an empty string or a plausible guess.
Keep that distinction when importing into another system. A missing opening-hours value does not mean closed; a missing review date does not mean checked today.
Dates and freshness
A review date records a check, not the time a file was built. Consumers should retain it and distinguish it from publication or download time.
The current site has no automatic check-log publication gate or fixed staleness threshold. Do not assume that a future export has either without reading and validating its own contract.
Checked on a date explains the limits of a review, while the method describes the editorial standard.
Licence and privacy
Do not assume that a reference specification grants rights to republish a dataset. Read the licence attached to the actual artifact. Business names, logos, quotations and third-party material may carry separate rights.
The reference is not a prospecting list. Do not add private contact information, infer personal details, or treat an unknown field as an invitation to enrich it.
Validation before use
- Confirm that the artifact actually exists and identify its publisher.
- Read its versioned schema and licence.
- Validate required fields and types.
- Preserve nulls, review dates and attribution.
- Reject or quarantine malformed records.
- Handle fetch failure and older snapshots explicitly.
Those are implementation requirements for a consumer you build. This article supplies neither an operating API nor a promise about service availability.
For corrections to a public listing, use corrections.
For the wider context, read Our Entry Bar, In Full: And Everything We Turn Down.
