How deals are classified

Two tests, applied in order, always from the data owner’s seat.

Why the seat matters. The same deal reads differently from each side. A dataset sold into someone else’s product is a sell for the owner and a wrap for whoever embeds it. This register always classifies from the position of the party that owns the data.

Two kinds of entry

The register records every real data deal it finds. Most go up as short entries: a summary of the public record, checked line by line against the sources before publication, with no classification and no reading. The ones marked ✓ Reviewed have been read and analysed by Gianluca: those carry the classification below, the reading and why the deal matters. A short entry can become a reviewed one later, at the same address.

Test 1

Who receives the value?

Internal — the asset makes the owner’s own operations or existing offer better, and no external party pays for it → improve. Stop here.
External — an external party pays → go to test 2.

Test 2

What is that party paying for?

The data, or an insight derived from it, is the deliverable — it is the line item → sell.
A product the data is inseparable from — the buyer cannot purchase the data without buying the product → wrap.

The cases that catch people out

Data staying internal does not make it improve. A diagnostic that reads images inside the hospital and returns a finding is a sell: the images never moved, but the finding is what was bought.
A subscription is not a wrap. Ask what is bought, not how often it is paid for.
Retail media is a wrap. Advertisers buy placements; the first-party data is inseparable from the inventory rather than purchasable on its own.
An acquisition priced on the data asset is a sell from the target’s seat. An acquisition of tooling or people, where no data asset changes hands, has no route at all.

The rake, and when it is “on loan”

The rake is the share of the value created that a data business keeps, deal after deal. It is owned where the owner does the enrichment. It is on loan where a partner does it: there is a rake, but it is contingent — if the partner walks, the value walks with them.

Where the buyer does the enrichment there is no downstream rake at all. The owner sold an input for a fee and the buyer keeps what it builds, permanently. A renewal in that arrangement re-prices the next licence; it does not recover the last one. The question worth asking there is whether the fee was under-priced — which is a sharper criticism, not a softer one.

Two kinds of non-answer

Unclear means there is a route but the public facts do not settle which. These entries are published anyway — 3 of them so far — a named deal marked unclear is more useful than a confident guess. No route means the event is not a monetisation event at all: a standards document, a capability acquisition. Those are reviewed and left out entirely.

What this register cannot show

Improve has no counterparty, so it produces no announcement. A register built from public disclosures structurally under-counts it. Nothing here is a complete picture of how data is monetised — only of how it is monetised in public.