both4 min read

Data licensing: what you are actually buying

Scope, duration, territory, exclusivity, derivatives. The five terms that decide the price of a dataset — and the one that causes almost every dispute.

4 min read ⏱

Data licensing, term by term

What you buy is a right of use — not a file

11 slides · swipe or use the arrows

d-nvest.com1/11

The starting point

You are not buying data. You are buying a RIGHT.

Data is not property in the way a machine is. What changes hands is a permission: to use, for a purpose, for a time, in a place. The file is only the delivery.

d-nvest.com2/11

Term 1

Scope — who may use it

Scope is where almost every dispute starts, because both sides assume the reading that suits them.

One team, one legal entity, or the whole group
Internal analysis only, or embedded in a product you sell
Human use, or training a model

d-nvest.com3/11

Term 2

Duration — a snapshot or a feed

A perpetual licence on a snapshot is not the same product as a one-year licence on a refreshed feed. Ask what happens on expiry: must you delete what you derived from it?

d-nvest.com4/11

Term 3

Territory — where the use happens

Rarely negotiated, occasionally fatal. A licence limited to the EEA does not cover a US subsidiary querying the same warehouse.

d-nvest.com5/11

Term 4

Exclusivity — the most expensive word

Sellers rarely grant full exclusivity: it kills every other deal on the same asset.

Non-exclusive — the norm, and the cheapest
Exclusive by sector or by geography — a middle ground
Fully exclusive — rare, and priced accordingly

d-nvest.com6/11

Term 5

Derivatives — what you may keep

If you train a model on licensed data, does the model outlive the licence? This is the term that AI made expensive, and the one most contracts written before 2023 simply do not address.

d-nvest.com7/11

The legal floor

In Europe, a database has its own right

The EU Database Directive protects a substantial investment in collecting or verifying data, independently of copyright. A dataset can therefore be protected even when no single record is.

┌ EU Directive 96/9/EC on the legal protection of databases

d-nvest.com8/11

Open licences

Free does not mean unconditional

Open data ends up in commercial products every day. The licence still decides whether that is allowed.

ODbL — share-alike: your derivative may have to stay open
CC BY — attribution required, commercial use allowed
CC BY-NC — no commercial use: incompatible with most products

d-nvest.com9/11

The trap

Personal data has a second lock

A licence grants CONTRACTUAL rights. It cannot create a legal basis under the GDPR. Resale is a new purpose — the seller must hold a basis that covers it, and a contract cannot manufacture one.

d-nvest.com10/11

Takeaway

Five questions before signing

A seller who cannot answer these has not licensed data before.

Who exactly may use it, and for what purpose?
For how long, and what must be deleted on expiry?
Where may the use happen?
May I build and keep derivatives, including models?
If personal data is involved: on what legal basis is it transferred?

d-nvest.com11/11

Browse available datasets

Questions about monetising or buying data?

Talk to an expert — no strings attached.

Book a free 30-min call

The full guide

Data licensing is misunderstood because the vocabulary of ownership does not fit. You are not buying data the way you buy a machine: you are buying a right of use, bounded by a purpose, a time and a place. The file is only the delivery mechanism. Five terms decide the price. Scope says who may use the data - one team, one entity, or the whole group - and whether it may be embedded in a product you sell or used to train a model. Scope causes almost every dispute, because each side assumes the reading that suits it. Duration distinguishes a perpetual licence on a snapshot from a one-year licence on a refreshed feed, and it raises the question everyone forgets: what must be deleted on expiry. Territory is rarely negotiated and occasionally fatal - a licence limited to the EEA does not cover a US subsidiary querying the same warehouse. Exclusivity is the most expensive word in the contract, and sellers rarely grant it in full because it kills every other deal on the same asset. Derivatives is the term that artificial intelligence made expensive: if you train a model on licensed data, does the model outlive the licence? Most contracts written before 2023 simply do not say. Underneath the contract sits a legal floor. In Europe, the Database Directive protects a substantial investment in collecting or verifying data, independently of copyright, so a dataset can be protected even when no single record is. Open licences are not exempt from care either: ODbL carries a share-alike obligation that can force your derivative to stay open, CC BY requires attribution, and CC BY-NC forbids the commercial use that most products depend on. Finally, the trap. A licence grants contractual rights; it cannot create a legal basis under the GDPR. Resale is a new purpose, and the seller must hold a basis that covers it. No contract can manufacture one. Before signing, ask five questions: who may use it and for what, for how long and what must be deleted, where the use may happen, whether derivatives may be kept, and - if personal data is involved - on what legal basis it is transferred. A seller who cannot answer has not licensed data before.