Licence terms

Data licensing agreements: questions to ask

A data licensing agreement should make the proposed exchange understandable: which material, which recipient, which uses and which obligations. Use these questions to brief your adviser once a buyer has made a specific request.

Last checked

Agree the boundaries of the deal before comparing prices. A confidentiality agreement is not a substitute for reviewing the proposed licence.

Start with a deal outline, not a downloaded contract

The UK Intellectual Property Office describes a licence as permission from a rights owner to another party to use protected intellectual property. It recommends professional legal advice before entering a contract. That is useful UK background, not a ready-made agreement for every dataset or jurisdiction.

An NDA concerns confidential sharing. Treat the confidentiality discussion and the proposed commercial permissions as separate questions for your adviser. Do not infer permission to train a model just because a recipient may evaluate information in confidence.

This article is a commercial briefing checklist, not contract wording. Northern Ireland, England and Wales, Scotland and the Republic of Ireland should not be treated as one legal jurisdiction. Ask an appropriately qualified adviser to identify the governing law and the rights involved. Start with the readiness checklist if you have not yet established what your company holds.

Describe the material and its version

Write a short delivery description before discussing a licence fee. Specify the collection, period covered, proposed format, exclusions and whether this is one extract or a continuing supply. Identify the version that a reviewer has seen. Avoid a vague description such as "all our operational data".

Ask how corrections, replacements and future records would be handled. Would an additional field or a new business unit require another agreement? Who decides whether a record falls inside the agreed scope? Put these questions beside the relevant draft wording, rather than relying on an explanation given in a meeting.

For an illustrative software archive, the discussion might concern selected issue descriptions and resolution categories, not the repository, attachments or customer messages. This is a hypothetical scope example, not a claim that these records are suitable or available. A data dictionary can help make the proposed fields clear without sending real rows.

Separate evaluation, training and onward use

Ask the proposed buyer to list what it wants to do. Does the request cover a limited evaluation, model training, internal research, a commercial product or onward distribution? Which named organisations would receive access, including service providers and group companies?

Make the same distinction for outputs. Ask what happens to annotations, transformed datasets, evaluation results and models developed using the material. Do not assume that an agreement's title answers those questions. Our training versus evaluation guide explains the different purposes; your adviser should examine how the actual wording treats them.

Record any uses your business will not consider. A clear refusal at this stage is more useful than an apparently attractive price attached to permissions you cannot grant. If the buyer cannot yet define the use, keep the discussion at description level.

Ask what exclusivity would prevent

If exclusivity is requested, ask the adviser to explain its practical effect on your business. Could you still use the material in your own product? Could you work with another buyer for a different purpose? Does the restriction concern one version, one sector, a territory or future records as well?

Write down the commercial opportunities you would be giving up and for how long. Do not assume a label such as "non-exclusive" settles every restriction elsewhere in the draft. Equally, avoid treating exclusivity as automatically unacceptable: the decision needs a defined scope and an assessment of your own plans.

Keep this discussion separate from valuation. Our licensing-value article covers the assumptions behind a price, not the legal effect of a clause.

Make acceptance and payment understandable

Ask for a written acceptance process before undertaking preparation work. What will be checked, by whom, using which criteria? Can the recipient reject only affected records or the whole delivery? What evidence will you receive, and what opportunity is there to correct a problem?

Then trace the payment steps. Distinguish an indicative range, an agreed fee, an invoice and cleared payment. Identify any dependence on another party's approval or purchase. Ask who pays for preparation if the review stops, and whether support, updates or repeat delivery are included.

As a practical exercise, have the person who would do the work read the delivery and acceptance sections. Ask them to describe the task in their own words. If they cannot tell when the job is finished, the commercial outline needs more work.

Work through the end of the arrangement

Ask what expiry or termination means for access, retained copies, backups and material already derived from the collection. Ask specifically about any model use the buyer proposes. Do not assume that ending access to a dataset reverses earlier use, or that every copy must disappear under terms you have not reviewed.

Bring your adviser a short list of unresolved points: ownership, third-party restrictions, confidentiality, personal data, liability and the practical route for a dispute. This checklist does not settle any of them. Keep a dated record of the draft and the answers so later revisions do not quietly change the commercial understanding.

If there is no concrete buyer request yet, do not commission a full agreement on speculation. Start with the buyer due-diligence questions and a non-confidential outline. Fairlode's first step is a description, not a file transfer or permission to use your records.

General information, not legal advice, a valuation or a buyer offer. Illustrative examples are not customer case studies.

Sources and further reading

Sources checked 16 October 2026. Historical research is background, not evidence of current buyer demand.

Put the questions into practice.

Start with a description. No files, customer records or source code.

Describe your recordsBack to the blog →