A qualified data owner is not a qualified dataset

A company can be a credible prospective supplier while a particular dataset remains unverified. Company qualification asks whether there is a responsible organization and a plausible path to a permitted transaction. Dataset qualification asks whether a defined contribution contains the evidence, coverage and quality required for a particular use. Both matter, and one does not substitute for the other.

Keep these decisions separate in procurement records. An introduction should tell the next reviewer what was checked, what was only reported and what remains unknown. It should not quietly turn an intake answer into a technical acceptance result.

What is the difference between source and dataset qualification?

Imagine a fictional company with a named commercial contact, a responsible systems lead and an account of several years of customer-support history. Its business context fits a research team's sourcing requirement. That is a reason to investigate the source.

It does not yet establish that the archive contains the required timestamps, that conversations survived a migration, that the contact controls every relevant right or that a prepared sample supports the research task. The company could be suitable to approach and the proposed dataset could still fail acceptance.

The opposite can happen too. A technically promising sample may arrive through a person whose authority is unclear. Useful records do not establish that this person can authorize disclosure or a license. The next action is to resolve authority, not to treat the sample's utility as permission.

Which claims are reported, demonstrated or accepted?

Use ordinary language that tells a reviewer how much reliance to place on each statement.

These states are a proposed internal vocabulary, not an industry standard. A statement should move between them only when new evidence supports the change. Record who made the statement, when, which scope it concerns and the document or observation supporting it.

An accepted sample is still a sample. If the full contribution differs in date range, processing or population, identify the extra review needed rather than inheriting acceptance automatically.

What can the sample establish about the wider archive?

Suppose a source demonstrates ten linked cases from a fictional archive. The buyer can see that the requested relationship exists in those cases. To understand how broadly it exists, the buyer also needs the selection method, eligible population and exclusions.

Ten cases chosen by a curator answer a different question from ten selected by a documented sampling rule. Neither description should be concealed. A curated batch can be useful for explaining the schema; a later representative review may be needed to assess coverage.

Keep counts tied to their denominators. If six of ten cases have an outcome, record that observation as six of ten reviewed cases. Do not call it a company-wide completeness rate. If two cases are excluded because their timestamps are unreadable, show the exclusions instead of quietly changing the denominator to eight.

Document whether the sample was transformed, which fields were removed and which links were reconstructed. A finding from the original records may not hold after preparation removes information required for the task.

Does knowing where data came from establish permission?

The W3C PROV overview describes information about the people, activities and entities involved in producing an item. A source record can use that idea to trace where a sample came from, which processing steps were applied and which version was reviewed.

That history is useful evidence. It does not by itself settle whether a new recipient may receive the records or use them for a proposed purpose. Keep the authority review alongside it, with a clear connection to the specific contribution and intended use.

A company operating a system is not necessarily the copyright owner of everything stored in it. Under 17 USC 201 and 202, US copyright initially belongs to the author subject to provisions including works made for hire, and ownership of a copy is distinct from copyright. Review employee and contractor arrangements and any transfers instead of assuming every commissioned contribution belongs to the company. Customer material, third-party components, confidentiality and privacy restrictions need their own checks; the copyright provisions do not clear a whole archive.

Avoid using a job title, NDA or willingness to participate as the entire authority record. Identify who is authorized to make the proposed decision and which supporting permissions or restrictions require review. If authority is unresolved, record the gap and limit the next step accordingly.

What should the dataset handoff tell the next reviewer?

A practical handoff should include the proposed contribution, intended use, evidence state for important claims, known exclusions and the next decision requested. It should distinguish commercial interest from technical and permissions decisions.

An example summary could read: "The source reports retained support history for the proposed period. The systems lead has confirmed the archive exists. Linked engineering outcomes have not been reviewed. The source has authorized this named recipient to receive the high-level profile for fit assessment; no dataset access is authorized. Next step: agree sample criteria and determine the appropriate permissions."

This is a fictional form of words. Do not reuse its authorization sentence unless the recorded permission actually covers that recipient and purpose. If the next recipient, scope or use changes, reassess the authority before the handoff occurs.

Datasheets for Datasets supports systematic documentation rather than informal assumptions about a dataset. Our handoff approach adds explicit evidence states and a next-decision field so reviewers can see which conclusions remain provisional.

Which acceptance decisions need separate records?

Track at least four decisions separately: source suitability, technical evidence, permission for the proposed use and commercial acceptance. Name the decision-maker for each. A team may combine these in its workflow, but the evidence should remain distinguishable.

This makes corrections manageable. If a migration gap appears, update the affected coverage claim and acceptance decision. If an authorized recipient changes, revisit the handoff permission. You should be able to explain what changed without pretending every earlier finding was wrong.

For related sourcing mechanics, see how to source enterprise workflow data and the data provenance guide. The buyer sourcing brief can record the required evidence and acceptance questions before a source review; it does not replace the separate authority decision. Where your team needs company sourcing or buyer-market research, share a high-level requirement. Avelence can support the agreed commercial work; the responsible parties still need to authorize disclosure and accept the actual dataset.

Frequently asked questions

Does a qualified source mean the data is ready?

No. Source qualification records checks about the organization, contact and plausible transaction path. Readiness concerns a specified dataset version and use, including retained links, coverage and acceptance criteria. A systems lead confirming that support history exists does not demonstrate that enough cases link to the outcomes required by the buyer’s task.

Can a sample prove the full archive is suitable?

A sample can support a broader conclusion only within the limits of its design and evidence. Record the eligible population, selection method, exclusions and preparation before interpreting the findings. Ten carefully chosen complete examples establish that those examples exist; they do not establish a company-wide completeness rate. A changed full delivery may need further review even after a sample passes.

Who should approve a dataset handoff?

The parties with authority over the material, recipient and proposed purpose should approve that step. Identify the exact artifact and version, intended recipient, permitted use and handling conditions. Commercial interest, an NDA or a technical acceptance result is not a substitute for this decision. Keep the authorization record linked to the handoff so a later scope change is visible.