The disappearance of third-party tracking has been discussed, for the better part of a decade, as an operational inconvenience. Identifiers go away, targeting gets harder, measurement degrades, and the industry looks for a replacement that restores the previous capability under a new name.
That framing has produced a great deal of engineering and very little strategy, because it treats as a technical problem something that is a change in the terms of the relationship.
What was actually being traded
The mechanism that is ending was built on a transaction the customer never knowingly entered. Data about a person's behaviour was collected in one context and used in another, by parties they had never encountered, to make decisions about them that they could not see. It worked for as long as it was invisible.
Regulation did not create the objection to this arrangement; it registered one that already existed. That distinction matters, because it determines what happens next. A rule imposed against a settled preference gets circumvented — the industry is very good at circumvention. A rule that codifies what people already wanted does not get circumvented, because each workaround produces the same reaction that produced the rule.
The pattern of the last few years supports the second reading. Every technically ingenious replacement for the third-party cookie — fingerprinting, probabilistic identity, data cooperatives — has met the same response, and the response has not weakened with familiarity.
Consent as an offer, not a checkbox
The productive reframing is to treat consent as part of the offer rather than as a gate in front of it.
Consider what a consent request actually is: a proposal to exchange information for something. Most implementations obscure both sides of that exchange — the customer cannot tell what they are giving or what they get — which is why the standard outcome is either reflexive refusal or unread acceptance. Neither produces a usable relationship.
An offer stated plainly performs differently, and the firms that have tested it report the same thing: fewer consents, materially better ones. Tell us what you are shopping for and we will stop showing you the rest is a proposition a person can evaluate. It also creates an obligation, because a customer who granted permission for a stated purpose will notice when the permission is used for another.
This is the part that makes it a governance problem rather than a marketing one. Honouring a stated purpose requires knowing, across every system, which data arrived under which promise — and most organisations cannot answer that question, because their data was accumulated in an era when nobody had to.
The advantage that is quietly forming
There is a competitive consequence that is easy to miss while everyone is discussing compliance.
Third-party data was, for twenty years, an equaliser. Any firm could buy roughly the same audiences as any other, which meant that targeting sophistication was not a durable advantage: it was a purchasable commodity available to competitors on identical terms.
Data given directly, for a stated purpose, by a customer who chose to give it, is not purchasable. It cannot be replicated by a competitor with a larger budget. It is, in the strict sense, a resource-based advantage — rare, valuable, and difficult to imitate — and it accrues fastest to firms that have something worth granting permission to.
The firms treating privacy as a compliance cost are optimising to minimise that cost. The firms treating it as an offer are accumulating an asset their competitors cannot buy. Ten years from now the difference between those two postures will not look like a difference in legal spending.
What to do on Monday
Three things, none of which require waiting for a standard to settle.
Know what you hold and under what promise it arrived. This is unglamorous, expensive, and the precondition for everything else.
Ask for less, and say why. A permission request that names its purpose and its benefit is a marketing message, and should be written by someone who can write.
Build the parts of measurement that do not depend on identity — geographic holdouts, staggered launches, aggregate modelling. They were always the more reliable instruments. Signal loss has merely removed the more comfortable alternative.