Suggested internal effort: 3–8 person-days for an initial metadata pass; assumption only. Export available catalogue fields and preserve their original provenance. Use IDs and counts by cohort, not an unexplained total file count.
The Datasheets for Datasets paper motivates transparent documentation of datasets. The owner-specific register below is an original commercial adaptation, not a reproduction of its questionnaire. Gebru et al. — Datasheets for Datasets
| Register field | Entry expected | Why the decision needs it |
|---|---|---|
| Cohort ID and item count | Source period / contract version / modality | Defines what the pilot covers |
| Coverage and freshness | Subjects, geography, dates, update cycle | Tests fit to a buyer's task |
| Components | Text, photographs, charts, audio, linked assets | Prevents silent inclusion of different rights |
| Rights evidence location | Agreement reference and reviewer | Makes the claim traceable |
| Restrictions and unknowns | Use, recipient, term, territory | Establishes exclusions and review workload |
| Technical state | Format, parse issues, duplicate estimate | Sizes preparation work |
| Existing monetisation | Products and outstanding commitments | Identifies conflicts and opportunity cost |
Keep “unknown” as a valid value. If an estimate is produced from a sample, record sample method and coverage. Do not convert inferred authorship, publication date or contract mapping into verified facts.
Who to contact: archive managers, acquisitions counsel and publishing operations. Completion condition: the team can identify a candidate cohort and the records needed to review it. The output is a scope map, not a certificate that all rights are clear.