Data manifest
By Gareth HoyleLast updated How we research
A data manifest is a structured inventory of a dataset that lists its sources, record counts, date ranges, formats and known exclusions.
A data manifest is a structured inventory of everything in a dataset: the systems it came from, record counts, date ranges, languages, file formats, sizes and what was removed. Buyers use it to price the data before they see a sample and to check the delivery against it afterward.
What goes in a data manifest?
| Field | What it records |
|---|---|
| Source system | The tool each record came from, such as Zendesk, Jira or Slack |
| Record count | Tickets, messages or files by type |
| Date range | First and last record dates |
| Languages | Each language and its share of records |
| Formats | JSON, CSV, Parquet or original files |
| Exclusions | What was removed, such as HR records or payment data |
| File hashes | A checksum per file for the delivered version |
Why do buyers trust a data manifest?
Buyers trust a data manifest when it is tied to a version and can be checked. Contributors to a 2026 Hugging Face forum thread on selling datasets suggested evidence packs bound to file hashes, so a buyer can confirm the delivered files match what was sampled.
Manifest accuracy is one of the eight points in our seller vetting. See how to build a data manifest and dataset card.