Brands didn’t get the cookie apocalypse they’d been bracing for. But the winners of 2025 act as if it happened anyway—shifting spend to contextual signals, retail media, and privacy‑safe data collaboration that out‑compete old‑school lookalikes on cost and incrementality.

The new baseline: user choice, not universal tracking

Google pivoted from a full deprecation of third‑party cookies to a “user‑choice” model in Chrome, while continuing to invest in Privacy Sandbox APIs. In practice, that means cookies persist for many users—but the direction of travel is unmistakable: more privacy controls, more platform variability, and less reliable cross‑site IDs.¹ ² Meanwhile, Safari and Firefox continue to block most third‑party cookies by default, so a large slice of traffic is already ID‑light.³ ⁴

Implication: Advertisers can’t bank on platform lookalikes trained on cross‑app or cross‑site data. The mixes that win now pair contextual reach with first‑party joins in clean rooms and retail media signals with closed‑loop measurement.

The mix that beats cookie‑era lookalikes

1) First‑party joins (PAIR/clean rooms) + contextual pre‑qualification
When an advertiser and a publisher reconcile their hashed first‑party data via a clean room—using standards like IAB Tech Lab’s PAIR protocol—and layer placements where the page itself signals intent, conversion rates rise and waste falls. PAIR’s cryptographic design (commutative encryption) enables audience overlap without exposing identities; early results have been strong.⁵ ⁶ ⁷

Evidence to watch: LiveRamp reported PAIR campaigns delivered 4× higher conversion rates than cookie‑based first‑party targeting in DV360, while Google cited +11% incremental reach for PAIR audiences vs. cookie lists.⁶ ⁷

• Why it wins: Contextual filters reduce off‑target impressions; the PAIR join lifts precision on users who actually know your brand—without third‑party IDs. And because both parties control their first‑party data, the setup is more resilient to consent and browser shifts.⁵
  
  

2) Retail media signals (on‑site and off‑site) + clean‑room measurement
Retail media networks (RMNs) bring durable purchase data and native closed‑loop attribution. In 2025 they’re expanding measurement windows and data access via clean rooms—e.g., Amazon Marketing Cloud (AMC) now supports up to five years of store signals.⁸ Retail media continues to gain share of digital budgets, with off‑site retail media (using RMN audiences across the open web) accounting for more than one‑fifth of spend.⁹ ¹⁰ 

• Evidence to watch: AMC‑guided optimization raised repeat purchases 20% in six weeks for one brand at flat spend.¹¹ While case studies vary, the pattern is consistent: when you plan with SKU‑level or basket data, CAC stabilizes even as third‑party IDs erode.
  
  
• Why it wins: RMN purchase graphs qualify intent better than pixel history, and clean‑room queries expose cross‑channel paths—what exposure mix natively drives the sale—without leaking user‑level data.⁸ ¹²
  
  

3) ID‑less cohorts + incrementality (geo‑lift/MMM) to steer scale
Cohort strategies (e.g., Curated Audiences—formerly Seller Defined Audiences) let publishers package first‑party signals without user IDs; IAB’s new ID‑less guidance formalizes techniques like k‑anonymity and differential privacy.¹³ ¹⁴ Pair these with incrementality testing to choose the lowest‑CAC contexts and creative. Across 190 geo‑based tests, YouTube’s incremental lift often far exceeded platform‑reported results—a reminder that post‑cookie channels require lift, not last‑click, to judge.¹⁵ ¹⁶ ¹⁷

How the numbers move: CPM, CAC, and closed‑loop proof

• CPM/CPC: Independent policy analysis aggregating vendor studies found contextual could deliver ~41% lower CPMs and ~48% lower CPCs than profiling‑based ads, albeit with caveats about limited independent replication.¹⁸ Vendor casework (Samsung/Mindshare + IAS) shows contextual outperforming standard platform targeting across CTR, cost per visit and cost per add‑to‑cart in APAC.¹⁹
  
  
• CAC: Clean‑room joins tend to improve qualifier density rather than blasting more reach. PAIR case studies (Google/LiveRamp) report +11% incremental reach at higher CVR, indicating better downstream CAC, while AMC lift cases show revenue gains at flat spend.⁷ ¹¹
• Incrementality: Peer‑reviewed and practitioner research converge: MMM + geo‑holdouts are the most reliable way to measure causal impact when IDs are sparse.¹⁶ ²⁰ ²¹ In practice, brands are re‑centering on lift, not last‑click, to judge contextual, RMN, and video. 

What the research says—and where it’s thin

There’s strong platform‑ and vendor‑provided evidence that first‑party joins (PAIR), retail media graphs, and contextual signals can match or beat cookie‑era tactics on conversion rate and reach quality.⁶ ⁷ ⁸ ¹¹ Independent meta‑analyses remain scarce; the most balanced reviews note promising economics for contextual but call out the need for more independent, large‑N studies that aren’t limited to single vendors or markets.¹⁸ Academic work continues to validate incrementality and MMM as decision frameworks in ID‑light environments.²⁰

Policy‑compliant plumbing: what “privacy‑safe” actually means

Modern clean rooms (ADH, AMC, Snowflake‑native, InfoSum) enforce aggregation thresholds and join rules so no user‑level data leaves the environment. Google’s Ads Data Hub requires ≥50 users for most query rows (≥10 for pure click/conversion), while AMC uses ≥100 users; results below thresholds are dropped.²² ²³ These systems, plus ID‑less specs and Chrome’s evolving controls, are why brands can safely run first‑party joins and still pass privacy muster.¹³ ²⁴