What the harvest takes¶

None of the assets here is a secret. They are the ordinary traces a life leaves while going about its business: a page opened, a route walked, a card tapped. What makes them valuable to the market is not that any one is sensitive but that they are continuous and, above all, joinable. The thing a person might want kept in its own context is exactly the thing the market prices out of it.

So the root asset is not any single trace. It is whatever lets the traces be tied to one person over time.

The join key¶

Strip the identifier and a person’s activity is a heap of scattered, unlinkable traces: a view here, a purchase there, none of it adding up. Restore it and the same heap composes into a profile. The identifier is the hinge on which the whole model turns.

Several kinds do the work. Cookie IDs in the browser, the advertising ID on a phone, device and browser fingerprints assembled from settings that seem harmless one at a time. The most durable is the hashed email address, because it survives a cleared cookie and a reinstalled app and follows a person from laptop to phone to television. This is the asset the deanonymisation model later turns back into a name. Here it does something quieter and more foundational: it makes everything else linkable.

The behavioural stream¶

What gets read, searched, watched, and bought, and how long attention rested on each. A single event is forgettable. The stream is a close account of attention and intent, kept in far more detail than memory would. Purchase records sit at the valuable end, because intent there is already realised rather than guessed: the thing was wanted enough to be paid for.

Location and pattern of life¶

Where a device goes, and when. Home and work fall out of the two places it rests, one overnight and one by day, without either being labelled. A clinic, a place of worship, a lawyer’s office, a border post each carry a meaning the bare coordinate does not announce. Continuous location is among the most identifying streams there is, because a person’s daily pattern is close to unique to them, which is part of why the broker market pays for it.

The inferred attribute¶

The highest-value asset is often one the person never handed over. From these same streams the market derives a sensitive trait and sells it as a ready-made audience. The FTC’s 2014 survey of the broker industry found segments carrying labels such as “Expectant Parent”, “Diabetes Interest”, and “Financially Challenged”. The label is the product. A buyer purchases the conclusion, not the evidence beneath it, and rarely asks how firm the guess was. Because the trait was calculated rather than collected, it slips past the intuition that sensitive data is the data a person chose to disclose. Nothing was disclosed. It was worked out.

The profile itself¶

The assembled dossier is an asset in its own right, worth more than the sum of the entries, because the value is made in the joining. It is valued, matched against other datasets, rented to bidders each time an ad loads, and sold onward. It is not a snapshot taken once and filed. It is a standing holding, topped up and re-priced as new traces arrive, and it keeps its worth long after the moment that produced any single line in it has passed.

Where the value concentrates¶

The streams are cheap and everywhere; the inference is clever; the profile is the saleable object. But all of it rests on the join key. Remove the persistent identifier and the rest falls back into unlinkable traces of no particular worth. Keep it and everything composes. Both the value to the collector and the exposure to the person gather at the same point, which is worth knowing before deciding where defence is best spent.

Last reviewed: 2026-07-16.