For about six years, the death of third-party cookies was a permanent industry talking point. Every six months a new deadline got announced, then quietly pushed back. Marketers learned to treat the warning the way Australians treat bushfire warnings in November—real eventually, but probably not this week.

Well, it actually happened. Chrome’s third-party cookie phase-out completed in late 2025. Most ad tech platforms transitioned through Q1 2026. The dust is starting to settle, and attribution looks different in ways the industry largely predicted and a few ways nobody did.

What Broke Worse Than Expected

Cross-domain retargeting attribution collapsed harder than most vendors admitted it would. The hand-wave answer for the last few years was that “alternative IDs” and “cohort-based” approaches would fill the gap. They have, partially, but the measurement quality is meaningfully worse than what cookie-based retargeting delivered when it was working.

The category most affected is what I’d call mid-funnel display—the prospecting-to-consideration retargeting that depended on tracking users across editorial sites, comparison sites, and review sites before they arrived at the advertiser’s domain. That whole pattern is largely gone for advertisers who haven’t built their own first-party data infrastructure.

Click attribution on programmatic display is still functional but the window has shortened dramatically. View-through attribution is essentially over for non-walled-garden display, and good riddance honestly. View-through measurement was always more of an art project than a measurement methodology.

Digiday published useful analysis in February showing programmatic display spending down approximately 22% year-over-year as advertisers shifted budget to channels with clearer attribution. The shift didn’t move proportionally to walled gardens—Meta and Google saw modest gains, not commensurate with display’s decline—suggesting that some budget left digital advertising entirely.

What Held Up Better Than Expected

Email attribution, of all things, has had a strong year. The combination of click-based tracking through email service providers and the comparatively healthy state of email deliverability has made email one of the easier channels to measure cleanly in a post-cookie world.

Connected TV attribution, after years of being aspirational, has actually started working. The platforms have invested in clean rooms and identity resolution that, while imperfect, deliver measurement quality comparable to where display was in 2019. For brands that committed early to CTV measurement, the post-cookie transition has been less painful than expected.

Server-side first-party tracking, deployed properly, has held up. The brands that invested in proper conversion API implementations, server-side tag managers, and first-party identity graphs over the last two years are in dramatically better positions than those who waited.

Direct attribution from search remains roughly where it was. Search clicks were always more attributable than display impressions, and that hasn’t changed. What has changed is that the relative value of search has gone up because nothing else in the digital mix is as measurable.

The Walled Gardens Won, Sort Of

The conventional wisdom that cookie deprecation would benefit walled gardens at the expense of the open web has largely played out, but the magnitude is smaller than predicted.

Meta’s measurement quality is better than the open web’s, but it’s also more opaque. Advertisers who interrogate Meta’s attribution carefully often come away unimpressed. Google’s measurement is similarly better-but-not-great. The walled gardens have benefited from the fact that they’re now the only places where reasonable cross-session tracking still exists, but they haven’t been able to translate that into the proportional spending shift the industry projected.

A more accurate description is that walled gardens have become unavoidable rather than dominant. Advertisers spend with them because they have to, not because they want to.

Modeling Has Become Mandatory

The change that’s affected my work most is that statistical modeling—media mix modeling, geo-experimentation, marketing experimentation generally—has gone from being a nice-to-have for sophisticated marketers to being a baseline expectation.

Five years ago, the brands running serious MMM programs were a small minority. Now, any client spending more than about $5 million annually who doesn’t have ongoing modeling capability is operating with measurement gaps that show up clearly in budget allocation decisions. The brands without modeling are over-allocating to lower-funnel channels because those are the ones still cleanly measurable. They’re under-investing in awareness and consideration channels because they can’t see the impact.

This is creating real competitive separation. Brands with modeling capability are correctly identifying when their attributable performance is actually understated incremental performance. Brands without modeling are flying blind.

The Tools Question

Vendor selection has gotten harder, not easier. The post-cookie landscape produced a wave of new attribution and measurement vendors, plus extensive rebranding from existing ones. Most of them are marketing the same three or four underlying methodologies with different terminology.

What I look for: transparent methodology documentation, willingness to participate in independent validation against geo-holdout tests, and an honest answer to “where does this approach break down?” Vendors who can’t answer the last question with specifics aren’t ready for serious work.

The other tool question is around clean rooms. Data clean room adoption has accelerated, but the use cases are narrower than the marketing suggests. They’re good for specific measurement questions where you have a media partner willing to share data structure. They’re not a general-purpose attribution solution.

What I’d Tell a CMO Right Now

If a CMO asked me what to do about post-cookie attribution today, the advice would be straightforward.

Get your first-party data infrastructure right. Server-side conversion tracking, a working customer data platform, and consistent user identity across your owned properties are the table stakes. Without these, nothing else in your measurement stack works well.

Build modeling capability, either in-house or through a serious partner. Reliance on platform-reported attribution as a primary decision-making input is a strategy that’s getting worse over time.

Reduce dependency on retargeting-heavy channels. If your performance numbers look great because of aggressive retargeting, expect those numbers to deteriorate as retargeting effectiveness continues to erode.

Run holdout tests regularly. Not annually—quarterly, at least, for any channel representing meaningful spend. The cost of misallocating budget based on poor measurement now exceeds the cost of testing for most advertisers.

The brands that adapt to this new measurement environment will compound their advantages over the next few years. The ones still pining for the cookie era will quietly fall behind.