# How to build a data manifest

URL: https://aidataforsale.com/learn/how-to-build-a-data-manifest
Publisher: AI Data For Sale (Marketing Signals Ltd)
Updated: 2026-11-10

A data manifest lists what a dataset contains, where it came from and the rights status of each part. A field-by-field template for sellers.

Author: Gareth Hoyle
Audience: Data holders

## In short

- A data manifest is an inventory of a dataset that a buyer can evaluate without seeing the data itself.
- It records systems, record types, volumes, date ranges, creators, rights basis and exclusions.
- A manifest lets a seller gauge buyer interest before exporting or sharing any records.
- Per-part rights flags let a buyer license the cleared portion while open questions are resolved.

## Manifest template

| Field | Example |
| --- | --- |
| Dataset name | Support tickets 2019 to 2025 |
| Source system | Zendesk |
| Record type | Ticket threads with resolution codes |
| Volume | 412,000 tickets, 2.1 million messages |
| Date range | January 2019 to June 2025 |
| Languages | English 94%, Spanish 6% |
| Created by | Customers and support staff |
| Rights basis | Privacy notice v4 (2018) permits service improvement; counsel review pending |
| Personal data | Yes, to be de-identified |
| Exclusions | Legal disputes, security incidents, enterprise accounts under NDA |
| Labels | Category, priority, resolution code, CSAT |
| Format on delivery | JSONL, one ticket per line |
| Status | Cleared, pending or excluded |

The numbers above are an illustration of format, not real data.

## Tips

- Count, do not estimate, where you can
- Note anything you are unsure about rather than leaving it out
- Version the manifest as rights questions are resolved

Send us your manifest, or a rough version of it, through [list your data](https://aidataforsale.com/list-your-data).

## Questions

### Do buyers need to see data before the manifest?

No. The manifest comes first; a sample follows only if there is real interest and rights are clear.

## Sources

- [Gebru et al.: Datasheets for Datasets (Communications of the ACM)](https://arxiv.org/abs/1803.09010)