Most conversations about the end of third-party cookies focused on targeting. The more consequential effect landed somewhere less visible: measurement. Targeting failures are obvious because campaigns underperform. Measurement drift is invisible, because the dashboard still fills in.
What actually happened
Two things that were expected to happen did not. Google announced in July 2024 that it would not deprecate third-party cookies in Chrome, choosing a user-choice mechanism instead. Then, in October 2025, it discontinued Privacy Sandbox altogether, retiring core APIs including Topics, Protected Audience and Attribution Reporting after low adoption and continued regulatory pressure.
Third-party cookies therefore remain functional for most Chrome users. Their reliability, however, has declined meaningfully — and reliability, not availability, is what measurement depends on.
The 47% problem
Average match rates on authenticated traffic have stabilised at around 47% across the open web, against roughly 68% in the cookie era.
Consider what that does to a conversion path. If roughly half of authenticated traffic resolves to a known identity, then the half that does not resolve is not attributed to the touchpoints that influenced it. Those conversions do not vanish from your revenue reporting — they land in direct or organic buckets instead.
The result is systematic, and it is directional rather than random: paid display is under-credited, and last-touch channels are over-credited. Any budget decision made by comparing channel ROAS across that boundary is comparing measurements taken with different instruments.
What the current stack looks like
Programmatic buyers now run a hybrid: authenticated identifiers such as UID2, ID5 and RampID, contextual signals, and clean-room measurement. Clean rooms matter here specifically because they permit matching without exchanging raw identifiers — the compromise that keeps measurement viable when the shared identifier layer is unreliable.
Meanwhile the market kept growing. Programmatic spend reached an estimated $755 billion in 2025 and is projected to clear $821 billion in 2026, about 90% of global digital display investment. The money did not leave; the visibility into it degraded.
Three corrections worth making
Stop comparing year-over-year attribution directly. A 2024 ROAS figure and a 2026 ROAS figure for the same channel were produced under different match conditions. The comparison looks valid and is not.
Run holdouts. Geographic or audience holdout tests measure incrementality without depending on identity resolution at all. They are more work than reading a dashboard, and they are the only method that did not degrade.
Treat first-party data as measurement infrastructure, not just targeting inventory. Publishers reached this conclusion first — 71% cited first-party data as a key driver of results in Q1 2025, rising from 64% a year earlier, with 85% expecting its role to grow.
The cookie deprecation that dominated planning for half a decade never happened. The measurement degradation nobody scheduled did. It is the quieter problem, and for anyone making budget decisions from attribution reports, the more expensive one.
