Verified empty inventory
A complete returned list may contain no items. An active PSC set may also be empty when every returned item is marked as ceased. These are different source conditions.
Data methodology & provenance
Healthcare market data with traceable source evidence.
See what the source reported, when it was observed, how recorded values changed, and where coverage remains incomplete.
CareDB connects Care Quality Commission (CQC) location and provider records with Companies House profiles, charges, officer inventories and Persons with Significant Control (PSC) lists where company links are available.
Evidence you can assess. Digests check retained evidence integrity. Separate audits check that recorded values agree with that evidence. Coverage and source limitations remain part of the picture.
Trace values to retained observations.
SHA-256 digests allow retained evidence to be checked for changes. New capture runs retain original response bodies and normalized payloads with separate digests. Historical evidence coverage and representation are documented; a normalized historical payload is not presented as an original network response.
Retained observations support direct source-field checks. Derived fields, national company portfolios and calculations are checked against their contributing inputs and calculation rules. They may draw on several observations rather than a single payload.
A digest does not establish that a regulatory filing is accurate, that an observation is current, or that a provider has been linked to the correct legal company.
Preserve changes as recorded versions.
Changed registration details, ratings and corporate facts are recorded through new historical versions, preserving prior business values. Recorded history reflects the observations we captured; it cannot reconstruct updates that were never observed.
Validated replay runs check that repeated ingestion preserves existing business history and avoids unnecessary versions. Corrections to projection rules may legitimately create new versions. Replay coverage is assessed for the retained evidence and population used in each run.
An unknown count is different from zero.
Unknown counts remain blank in CSV outputs. Verified empty inventories produce zero. Missing evidence and failed requests are not interpreted as empty lists.
Exemption or non-standard structure
Explicit exemption metadata is retained where available in the source payload. A non-standard legal structure requires its own qualification; it is not inferred from an empty response.
Unavailable response · HTTP 404
A captured 404 records what the source returned. It does not establish exemption, dissolution, deregistration or an empty inventory, and is tracked separately from a missing capture.
Transient disruption
Failed requests retain available error and retry metadata. Missing or unavailable evidence is not interpreted as a verified empty inventory.
Missing evidence
Where no usable capture is retained, the inventory remains unknown. An endpoint that has not been requested is not classified as a network failure.
Check the projection as well as the payload.
For assessed release candidates, independent audits compare generated fields with captured inputs across the specified population. Checks cover field selection, rating dates, national portfolios, ISO timestamps, calculations and zero/null meanings without relying on production runtime classifiers.
CQC location [LOCATION_ID] → linked company [COMPANY_NUMBER]
Recorded version interval
valid_from[TIMESTAMP_UTC]valid_to- NULL · open interval
Retained source evidence
- Observation instant
[TIMESTAMP_UTC]- Raw response digest
- SHA-256
[RAW_RESPONSE_DIGEST] - Response code
- HTTP 200
Corporate inventories · separate observations
- Officer inventory
- Verified empty: zero recorded officer identities in the complete returned list at its observation time.
- PSC inventory
- Two active identities in the returned set captured at
[PSC_OBSERVATION_UTC].
Matching a projection to retained evidence establishes consistency within that assessment. It does not certify every historical record, earlier customer file or upstream fact.
Keep four kinds of time distinct.
Version intervals, source-declared dates, observation times and processing times answer different questions.
| Time | What it describes |
|---|---|
Version intervalvalid_from / valid_to | The recorded version’s interval in the database. It is not necessarily the regulator’s effective date. |
| Source-declared date | Dates or timestamps supplied by the regulator, such as registration dates or published rating dates, retained as distinct attributes. |
Observation timelast_source_observed_at | The UTC instant of a source capture where retained. Historical dated caches may instead have documented observation-window bounds. |
Processing timecreated_at / updated_at | The internal database write or processing time, according to the table’s timestamp meaning. |
Ingestion and export database paths use explicit UTC handling, tested across timezones and daylight-saving boundaries for consistent timestamp serialization. Source calendar dates remain calendar dates.
Know what the evidence can establish.
- Evidence integrity. Digests check retained content. They do not rule out source filing errors, stale entries or incorrect company associations.
- Point-in-time observations. Capture dates and documented cache windows describe when evidence was observed. A snapshot time is the database read time, not a claim that every upstream record was fetched then. Freshness and available age information depend on the assessed product and evidence coverage.
- Historical gaps. Retained evidence covers documented observation windows. Changes that were never captured cannot be reconstructed from a current source response.
- Corporate relationships. Recorded officer identities include represented appointments, including resigned positions where captured, and do not represent current board membership or headcount. Active PSC counts reflect regulatory filing lists and do not determine ultimate beneficial ownership. A company association alone does not establish ownership.
Bring the source context into your research.
Explore the product fields and examples, or contact us to discuss the coverage and evidence relevant to your workflow.