Few things in this industry have absorbed as much planning effort as the deprecation of third-party cookies in the largest browser. It was announced, postponed repeatedly, and then the approach changed materially.
A lot of people concluded from that they had wasted their preparation. I think the opposite, and the reason is that the deadline was never the real change.
What had already happened before any of it
The thing that gets lost in the focus on one browser.
Other major browsers had already restricted third-party cookies years earlier, by default, affecting a substantial share of traffic in many markets.
Mobile operating system changes had already made cross-app identification opt-in, and opt-in rates were low enough to change the economics of app advertising significantly.
Privacy regulation in multiple jurisdictions had already imposed consent requirements that reduced the population of trackable users regardless of browser behaviour.
So by the time the headline deprecation was being debated, a large fraction of signal loss had already occurred through other routes. Organisations that treated it as a future event were describing something that had partly happened.
What the preparation actually produced
This is why I do not think it was wasted.
First-party data collection improved across the industry, because organisations finally had a reason to prioritise it. That is durable value independent of any browser.
Consent management became a solved problem operationally rather than a compliance afterthought.
Measurement approaches that do not depend on individual tracking — incrementality testing, media mix modelling, geographic holdouts — came back into use after a decade of neglect.
And a lot of organisations discovered how much of their targeting was contributing nothing, because the tests they ran to prepare for signal loss revealed that removing certain segments changed very little.
The alternatives and where they landed
A fair summary of the landscape, without predicting which wins.
Contextual targeting returned, substantially improved from its earlier form, because the technology for understanding page content has advanced considerably. It requires no user data at all, which makes it durable.
Publisher first-party data and direct relationships gained value, which advantages large publishers with logged-in audiences and disadvantages everybody else.
Clean rooms and data collaboration environments emerged for matching data between parties without either seeing the other's raw records. These work and they are operationally heavy.
Identity solutions based on hashed email and similar identifiers achieved partial adoption, with the persistent problem that they only cover users who log in somewhere.
And the browser-native proposals achieved limited traction, partly for technical reasons and partly because the industry did not trust the party proposing them.
What I would tell somebody planning now
Do not plan around a date. The dates have moved repeatedly and the direction has not.
The direction is that individual-level tracking across sites is becoming harder, less complete, and more legally constrained, through many independent mechanisms. That trend does not reverse because one browser changes its mind.
Which means the useful investments are the ones that hold regardless: your own data, direct relationships, contextual capability, and measurement that does not require observing individuals.
All of those are valuable if nothing else changes, which is the test I would apply to any preparation.
The measurement consequence specifically
The part that has changed working practice most in the places I have seen.
Deterministic measurement of individual journeys is no longer available for a large share of activity, and the gap is filled by modelling.
Which means numbers on platform dashboards are increasingly estimates produced by interested parties, and the appropriate response is to treat them as directional and to establish an independent source of truth.
For most organisations that means a controlled test capability — the ability to hold out a region or an audience and compare — which is unglamorous, disliked by finance, and the only method that does not degrade as tracking degrades.
The thing that has not changed
Worth ending on because the panic obscured it.
Advertising worked before any of this existed. Television, print and radio were bought and measured for decades without individual tracking, using panels, models and tests.
The techniques for doing that were not lost, they were set aside because something more precise appeared to be available. A good deal of the current work is rediscovery rather than invention.
Which is a reasonably encouraging position to be in, and considerably less dramatic than the coverage suggested.
What to do with a plan already written
For teams that built a deprecation roadmap and then watched the deadline move, the useful exercise is to sort it rather than shelve it.
Anything that improves first-party data collection, consent handling or measurement independence is worth doing regardless and should proceed on its own merits.
Anything built specifically around one browser's proposed replacement mechanism should be paused, because those proposals have changed repeatedly.
And anything justified purely by a compliance deadline should be re-justified commercially, because if there is no case beyond the deadline there was probably no case.
That sorting takes an afternoon and converts a stalled roadmap into a shorter list of things worth funding.