MMM is the cross-channel anchor again.

Last-touch attribution depended on one continuous identifier following a person from impression to conversion. That identifier is gone. On October 17, 2025, Google retired ten Privacy Sandbox technologies, citing low levels of adoption; only CHIPS, FedCM, and Private State Tokens survive, and third-party cookies remain in Chrome with no choice prompt (Google, October 2025). The UK's CMA released Google from its commitments the same day (MediaPath Research, The State of Digital Media 2026, citing reporting on CMA compliance documents, 2025). The reversal changed less than it appeared to. The post-cookie identity stack of UID2, ID5, LiveRamp ATS, and RampID kept building straight through it, because the rebuild answers a structural problem the regulator's decision never touched (MediaPath Research, The State of Digital Media 2026).

The replacement plumbing never matched the cookie's precision. Reporting on Google's compliance documents to the CMA found that up to 85% of conversions tracked by the Attribution Reporting API showed 60 to 100% inaccuracy against cookie-based measurement (MediaPath Research, The State of Digital Media 2026, citing CMA compliance documents, 2025). The technical inadequacy of the Sandbox alternatives is what drove the deprecation. What replaced the cookie is a stack of partial identifiers, each covering a different slice of the consumer's day.

Audiences moved while the plumbing fractured. Brands now run six to nine channels by default, and identity, signals, and creative travel with the viewer (MediaPath Research, The State of Digital Media 2026). The channel labels a click path relies on are dissolving underneath it. Off-site retail media, which runs a retailer's shopper data against inventory the retailer does not own, grows 42.1% year over year against 15.3% on-site and now accounts for roughly 20% of total retail media spend (eMarketer, 2026). YouTube took 13.4% of US TV viewing time in July 2025, the largest media-distributor lead since the Nielsen Gauge launched (Nielsen, July 2025). A retail media dollar now lands on a television, and the biggest single thing on that television is a social platform. MTA fragments at exactly those boundaries, and walled-garden reporting stays trapped inside its own gardens (MediaPath Research, The State of Digital Media 2026). Multiplied across a six-to-nine-channel plan, the attribution report stops describing what happened.

The ANA's Q2 2025 programmatic supply chain transparency benchmark surfaced $26.8 billion in annual programmatic waste (ANA/TAG TrustNet, Q2 2025). The same study documented how far the supply chain has already tightened. Made-for-advertising spend, which stood at 15% of programmatic in the study's 2023 baseline, fell to a 0.8% median by Q2 2025. Private marketplace share climbed from 64.5% to 87.8% across the same window (ANA/TAG TrustNet, Q2 2025).

The working-media arithmetic moved with the cleanup. $439 of every $1,000 entering a DSP now reaches consumers, up 7.9 percentage points against the study's earlier baseline (ANA/TAG TrustNet, Q2 2025). More of the surviving spend moves through private marketplaces and disciplined supply paths (MediaPath Research, The State of Digital Media 2026).

The cleanup sharpened the measurement question rather than settling it. When waste dominated the conversation, finding the leaks was the job. With MFA at a 0.8% median and private marketplaces carrying 87.8% of spend (ANA/TAG TrustNet, Q2 2025), the live question becomes what the surviving dollars cause across every channel they touch. Last-touch cannot answer that across channel boundaries, and the same 2025 research cycle that quantified the cleanup also recorded the methodology shift that follows from it.

Asked which methodology they rate most reliable for cross-channel measurement, US marketers gave marketing mix modeling a 46.9% share, incrementality testing 34.1%, and last-click multi-touch attribution 19.4% (eMarketer/TransUnion, July 2025). The methodology that requires no user-level identifier leads the ranking. The one built entirely on click paths sits last.

That puts MMM back at the center of cross-channel planning for the first time since 2008 (MediaPath Research, The State of Digital Media 2026). The timing follows from the mechanics. Identity fragmentation is what broke last-touch attribution, and it left MMM untouched: the method reads aggregate effect and never needed a user-level identifier to do it. Walled-garden attribution stays mature inside each platform, and comparing across platforms still requires MMM (MediaPath Research, The State of Digital Media 2026).

The vendor stack is named and partly open. Modern MMM runs on Recast, Analytic Partners, Nielsen Compass, Magic Numbers, and Meta's open-source Robyn (eMarketer/TransUnion, July 2025). Open-source models widened access to a method (MediaPath Research, The State of Digital Media 2026).

The IAB and MRC shipped Attention Measurement Guidelines v1.0 in November 2025, giving attention metrics from Adelaide, Lumen Research, TVision, Realeyes, and Amplified Intelligence a common standard (IAB/MRC, November 2025). Attention now has a definition buyers can plan against. A standard is the precondition for using a metric at the brief stage, and v1.0 makes attention defensible there (MediaPath Research, The State of Digital Media 2026).

Four dates in 2025 point the same way. The ANA benchmark quantified the cleanup in Q2. eMarketer and TransUnion recorded the reliability ranking in July. Google retired most of the Privacy Sandbox in October. The IAB and MRC standardized attention in November. Each pushed planning toward methodologies that survive identity fragmentation and compare across channels (ANA/TAG TrustNet, Q2 2025; eMarketer/TransUnion, July 2025; Google, October 2025; IAB/MRC, November 2025).

The planning consequence is sequencing. Set the outcome-measurement architecture before deciding which channels carry which dollars, with MMM, incrementality, and attention as the frame (MediaPath Research, The State of Digital Media 2026). The anchor methodology belongs at the brief stage, where it shapes allocation before the first dollar moves. The report's guidance on last-touch MTA is to bypass it.

Identity work runs in parallel. UID2, ID5, LiveRamp ATS, and first-party data clean rooms sit in the plan as foundational infrastructure (MediaPath Research, The State of Digital Media 2026). Chrome keeping the cookie changed none of that scheduling; the report reads the cookie's survival as an accelerant to the rebuild. Activation still runs on identifiers, and cross-channel proof now anchors on the model.

The same order holds in practice: MMM as the cross-channel anchor, incrementality as the validation layer, attention as a brief-stage input, with the measurement frame set out ahead of channel allocation by default. The methodology earned the rebound because it never needed the identifier that broke. For a 2026 plan running six to nine channels, that is the anchor every channel gets measured against.