The shift from MTA to macro-measurement

1. What multi-touch attribution promised

   MTA stitches one shopper's ad clicks and views into a path, then splits credit for the order across the touches: by rule (linear, time decay) or by statistics (data-driven). Every version needs to recognize the same person at each touch, through a cookie, a device ID, or a login.

   It had problems before privacy rules arrived. Each walled garden reports only the conversions it saw. Two platforms can claim the same order, so their reported revenue adds up to more than the store sold. View-through credit rewards an ad someone saw, whether or not it changed anything. And a path shows who was exposed, not what the ad caused: in 15 large experiments at Facebook, common observational methods “often fail to accurately measure the true effect of advertising” when checked against randomized tests.

   SourcesGordon et al., Marketing Science, 2019 (opens in a new tab)

2. What broke it

   Between 2020 and 2022, Safari, Apple's tracking prompt, and Firefox removed much of the identity MTA depends on. The platforms adjusted their own reporting to match.

   1. March 2020Safari blocks third-party cookies by default: “Cookies for cross-site resources are now blocked by default across the board.” (WebKit (opens in a new tab))
   2. January 2021Meta tells advertisers that once Apple enforces its prompt, web optimization is limited to 8 conversion events per domain, the longer click and view windows go away, and “statistical modeling will be used” for some metrics. (Meta for Developers (opens in a new tab))
   3. 26 April 2021iOS 14.5 ships App Tracking Transparency. Apps must “get the user's permission before tracking their data across apps or websites owned by other companies.” (Apple (opens in a new tab))
   4. 14 June 2022Firefox turns on Total Cookie Protection for all users: a separate cookie jar for each site. (Mozilla (opens in a new tab))
   5. October 2023Google removes first click, linear, time decay, and position-based models from Google Ads and GA4. Data-driven and last click remain. (Search Engine Land (opens in a new tab))

3. Chrome kept its cookies

   Chrome is the exception, and it is often misreported. In April 2025 Google said it would not roll out a standalone cookie prompt. In October 2025 it retired most Privacy Sandbox APIs, including Topics and the Attribution Reporting API, while keeping third-party cookie choice in Chrome settings.

   Cookies in Chrome do not bring back an iPhone user who declined tracking, or a Safari visit.

   SourcesPrivacy Sandbox, April 2025 (opens in a new tab)Privacy Sandbox, October 2025 (opens in a new tab)

4. What replaced it

   Two things filled the gap. The platforms began modeling what they can no longer see, and advertisers moved toward macro-measurement: methods that read totals, such as daily sales against spend by channel, instead of one shopper's path.

   Modeled conversions. Google says modeled conversions “use data that doesn't identify individual users to estimate conversions that Google is unable to observe directly,” and names Safari, Firefox, and consent rules as reasons. Platform ROAS is part count, part model.

   Marketing mix modeling. MMM regresses sales on spend by channel over time, with controls for seasonality and price. It needs no user-level data, and it has known limits. Google made its open-source Bayesian MMM, Meridian, available to everyone on 29 January 2025, with support for calibrating it to experiment results.

   Incrementality and lift tests. Turn spend off, or up, in some regions or for a holdout group, and compare against the rest. Google researchers described geo experiments for this in 2011. A test answers one question well, once.

   Blended ROAS (MER). Total sales divided by total marketing spend. No model, no overlap, every order counted once. It does not say which channel did the work.

   SourcesGoogle Ads Help (opens in a new tab)Chan and Perry, Google, 2017 (opens in a new tab)Google: Meridian (opens in a new tab)Google Research, 2011 (opens in a new tab)

5. What it means for the budget

   Platform reporting still earns its place inside one channel, for creative and campaign choices. It is weaker as the source of truth for the total budget.

   For the total, I read totals: MER for the cash check, MMM for where each channel's returns flatten, and a lift test when two channels claim the same sale.

Originally published at Mcfly Analytics (opens in a new tab) · The full article, with every source and a Q&A.

Try it: MMM budget reallocationTry it: Geo test readoutTry it: Executive MER dashboardHow I'd approach an MMMHow I'd design a geo testHow I'd reconcile platform numbers with cash