Life After ATT: Measuring What You Cannot See

It has been over a year since App Tracking Transparency landed, and the industry has moved through denial, anger, and bargaining. Opt-in rates settled far below what platform-side attribution needs to function, Meta publicly attributed a ten-figure revenue hit to signal loss, and every performance dashboard we inherit from a new client still quietly pretends none of this happened.

Time for the acceptance phase. Deterministic user-level attribution on iOS is gone, it is not coming back, and the teams still optimizing to last-click platform numbers are optimizing to fiction. Here is what we run instead.

Stop grading the platforms with their own homework

The core problem: ad platforms now model a large share of the conversions they report. Modeled conversions are not fake, but they are the platform's estimate of its own effectiveness, which is roughly like letting a contractor write their own inspection report.

Our rule with clients is simple. Platform-reported numbers are for in-platform optimization, comparing campaign A to campaign B inside the same walled garden. They are never for budget allocation across channels. The moment you use Meta's dashboard to decide Meta's budget against Google's dashboard deciding Google's budget, you have guaranteed over-investment in whichever platform models most aggressively.

The triangulation stack

No single method replaces what we lost, so we triangulate three imperfect ones:

• Marketing mix modeling, lightweight edition. Not the six-figure consultancy version. A regression on two-plus years of weekly spend and revenue data, refreshed quarterly. Crude, but it catches the big lies. For Northwind, it surfaced that paid social was claiming roughly double the revenue the model could actually find.
• Incrementality tests. Geo holdouts remain the most honest instrument in the toolbox. Turn a channel off in matched regions, measure what actually disappears. Painful to sell internally, worth every awkward meeting. Run two or three per year on your biggest lines.
• First-party data and self-reported attribution. A simple "how did you hear about us?" field at checkout is embarrassing technology and startlingly useful directionally, especially for channels attribution never credited well: podcasts, influencers, dark social.

Change what "performance" means

The deeper shift is cultural. The 2016-2021 era trained a generation of marketers to expect real-time, penny-accurate ROAS, and that expectation is now the biggest obstacle to good decisions. Weekly certainty was always partly illusion. The illusion just got harder to maintain.

We push clients toward a slower cadence: monthly reads, quarterly reallocations, annual strategy. Fewer numbers, more confidence in each. And we push spend toward the things signal loss cannot break: creative quality, brand memorability, and owned channels. Email and SMS lists have quietly become the highest-leverage assets in most of our clients' stacks precisely because no OS update can take them away.

The uncomfortable upside

Here is the contrarian take we have come to believe. ATT was good for marketing, if not for marketers' comfort. The old precision rewarded whoever could most efficiently harvest existing demand, and it starved the work that creates demand. Now that the harvesting math is blurry, the case for brand, creative, and genuine differentiation is back on the table with finance teams.

The winners of the next few years will not be the teams that reconstruct the old tracking by other means. They will be the teams that got comfortable making bigger decisions with honest uncertainty, sooner than their competitors did.