Key takeaway
Describe what the records must explain and how you will judge them. A generic request for “all company data” makes both evaluation and permissions harder.
Define the intended use and unit of evidence.
Specify whether the requirement is for training, evaluation, retrieval or another purpose. Name the workflow, language, period and relationships you need. If verified outcomes matter, explain what counts as verification and which unresolved cases remain useful.
Ask for provenance before volume.
Understand who created records, how they were captured and which systems changed over time. Request category-level descriptions of rights and exclusions first. A large archive with unclear source authority may require substantial work before a sample can be considered.
Agree on a fair evaluation.
State the minimum sample scope, success criteria, review period and the reasons for rejection. Ask whether a schema or illustrative example can answer the first question. Identify who has access and prohibit unintended use during evaluation where appropriate.
- Field meanings, coverage and relationship continuity
- Rights evidence, exclusions and company authority
- Disclosure controls and proposed retention
- Quality limitations and transformation history
- Acceptance criteria and a correction process
Make the commercial route explicit.
Clarify price, payment trigger, exclusivity, permitted recipients and refresh requirements. Explain the treatment of raw records, derived material and trained models. Keep data quality warranties proportionate to the source company’s actual knowledge and control.
Respect the company’s approval boundary.
An interested owner may permit a conversation without permitting record disclosure. Use a named recipient and exact scope at each handoff. VOID can discuss sourcing requirements and explore permissioned introductions; a buyer inquiry does not establish inventory, an exclusive seller mandate or an existing program relationship.