A diagnostics dataset and the limits the firm placed on it

In the third quarter of 2024 the Information & Data division took a 24 per cent position in a network of veterinary diagnostic laboratories operating across four countries. The commercial asset is not the laboratories. It is eleven years of test results, roughly 31 million records, linked to species, region, season and treatment outcome. A record of that length shows how a disease pattern moves across a region before any single practice notices. The question put by the Information Governance Committee was not what the data could show. It was what the firm was entitled to do with it.

Provenance of data is the third principle of the Ethical Technology Charter, and it is the one that costs money. Every dataset acquired must have a documented origin and a lawful basis. The seller presented nine sources. Seven had consent or contractual terms that could be evidenced back to the practice that submitted the sample. Two could not, because they had been acquired in an earlier transaction whose records were incomplete. Those two were excluded from the transaction perimeter and from the model that valued it.

The exclusion reduced the price the firm was prepared to pay by 18 per cent against the seller's own model. That was not a negotiating position. A dataset whose basis cannot be evidenced cannot be used, cannot be sold on, and carries a remediation cost that falls on whoever holds it when the question is finally asked. The seller accepted the reduced price after a second bidder reached the same conclusion. A buyer who pays for unevidenced data is paying for a liability with a delivery date it does not control.

Restraint in inference set the operating limits. The Charter permits no inference about individuals beyond what a mandate requires. In practice that meant three conditions written into the shareholders' agreement. Analysis is conducted at herd, practice or regional level and never at the level of an identified owner. Identifiers are removed at the point of collection rather than at the point of analysis. Results are not sold to insurers or to anyone pricing an individual policy. The third condition removed a revenue line the seller had projected.

What remains is the use the firm bought it for. Seasonal and regional disease patterns inform the underwriting of animal-health positions and the pricing of livestock and aquaculture facilities in the Commercial Finance division. A rise in a particular pathogen across a region is a credit signal two quarters before it reaches a borrower's mortality figures. That signal belongs to the firm's own underwriting. It is not a product, and the prohibition on selling client data extends to it as a matter of policy.

A named person answers for the arrangement. The Director of Data Governance in Toronto holds the register of sources, the lawful basis for each, and the date of the last verification. The Information Governance Committee reviewed that register in June 2024 and confirmed that all seven retained sources remained evidenced. The register is dull, and it is the whole control. An organisation that cannot say where a dataset came from will be asked eventually, and the answer does not improve with time.

Published 2024-10-15 by the Information & Data division. Research is prepared for eligible counterparties and does not constitute advice.