Five years on, marketers are still quoting the wrong number. The familiar "15-25% ATT opt-in" statistic comes from 2021-2022, and it survived because it fit the industry's preferred narrative: users rejected tracking, attribution broke, and performance marketing became less measurable. The data has changed. According to Adjust's Q1 2026 benchmark, global opt-in has recovered to roughly 38%, up from 35% a year earlier, continuing a steady upward trend as apps became better at explaining the value exchange behind permission requests.
For European brands, however, the picture is less optimistic. Opt-in rates remain around 22% across the EU, compared with roughly 31% in the US, leaving DACH advertisers with materially less deterministic signal to work with. At the same time, the regulatory picture has kept moving on two fronts. France's Autorité de la concurrence fined Apple €150 million over ATT in March 2025 - the first antitrust penalty anywhere specifically targeting the framework - finding it unfairly disadvantaged smaller publishers and rivals. Germany's own Bundeskartellamt opened a parallel case the same year, charging Apple with giving its own apps preferential treatment under the same system. Privacy is no longer just a technical constraint for European brands. It's an active legal battleground on two fronts.
But ATT was never the real story. It simply exposed a measurement problem that had existed for years.
The illusion was already there
Long before ATT, platform attribution was systematically over-crediting itself.
Geo-holdout experiments from the past two years consistently reach the same conclusion. In one Common Thread Collective (CTC) study, a premium fashion retailer paused Google branded search across test regions. Platform reporting predicted catastrophic losses. The actual result was seven fewer orders. Those campaigns had been claiming roughly five times more value than they were creating. The brand ultimately reduced branded search spend by 84%, cut overall media investment by 46%, and retained 99% of total orders.
Similar patterns appear elsewhere. Stella's 2025 benchmarks estimate that platform-reported ROAS often exceeds true incremental ROAS by 2-3×, while Haus found Meta retargeting commonly overstates performance by 40-70%. A typical example illustrates the gap:
| Metric | Platform report | Incrementality result |
|---|---|---|
| Meta retargeting spend | €50,000 | €50,000 |
| Attributed revenue | €380,000 | €90,000 |
| ROAS | 7.6× | 1.8× |
Meta and Google are not fabricating revenue. They're answering a different question. Attribution asks which ad was seen before the purchase. Incrementality asks whether the purchase would have happened anyway. Those are fundamentally different measurements, and confusing them is what inflated ROAS for a decade.
What actually replaced attribution
The post-ATT measurement stack isn't a new tool. It's two complementary disciplines.
Marketing Mix Modeling (MMM) estimates channel contribution using aggregated data - media spend, seasonality, promotions, pricing, and revenue over time - rather than following individuals. Triple Whale's Moby, Northbeam, Recast, Measured, and Haus all approach this differently, but share the same premise: causality emerges from patterns, not user-level tracking.
Incrementality testing does the opposite. Instead of modeling everything continuously, it isolates one question through controlled experiments - pausing Meta in selected regions, holding out branded search, running ghost ads - to measure what actually changes.
| MMM | Incrementality | |
|---|---|---|
| Method | Continuous estimation | Ground-truth experiments |
| Best for | Budget allocation, long-term planning | Validating one channel at a time |
One predicts. The other verifies. That distinction is reflected in how executives now evaluate evidence. Enalitica's 2026 research found 60% of senior decision-makers place the highest trust in independent incrementality tests, while 40% rank MMM highest. Almost nobody considers platform dashboards sufficient on their own.
The dashboard wasn't replaced. It was demoted.
The new operating model
This changes how high-performing DTC teams make budget decisions.
A platform may report 4.2× ROAS. An MMM model might estimate that Meta generated 35% of incremental revenue, with a confidence interval of ±8%. At first glance, the second number looks worse - less precise, less satisfying, harder to present in a board meeting. It's also considerably more useful. False precision encourages overconfidence. Honest uncertainty tells you how aggressively you should bet next quarter's budget.
For lean DTC teams, the practical implications are straightforward: watch blended CAC and MER first, since total spend against total revenue is far harder to manipulate than channel-level attribution; use platform ROAS for optimization, not strategy - valuable for creative testing and bidding, weak evidence for reallocating budget; and run one incrementality experiment each quarter, testing the two largest channels, which is usually enough to recalibrate every other number in the reporting stack.
You don't need an enterprise measurement budget to adopt this discipline. Brands spending under €1 million annually can already access lightweight MMM platforms and affordable incrementality tools. The competitive advantage isn't buying more software. It's changing which evidence deserves the final word.
The pattern, again
This is the same lesson that appeared in the CRM conversation, just in a different form. For years, deterministic attribution allowed marketers to substitute tracking for causality. The pixel appeared to answer the hardest question in marketing, so very few teams asked whether it actually did.
ATT didn't destroy measurement. It removed the illusion that measurement had already been solved.
What replaces a dashboard that lies with confidence isn't another dashboard that promises certainty. It's a decision-making discipline that accepts imperfect information, measures causality wherever possible, and is willing to act despite the remaining uncertainty.
Sources
Adjust, Global ATT Opt-In Benchmark (Q1 2026) · Digital Applied, "Mobile App Marketing Statistics 2026" · Autorité de la concurrence, Apple ATT decision (March 2025) · Bundeskartellamt, Apple ATT proceeding (2025) · Common Thread Collective (CTC), DTC geo-holdout case studies (2026) · Stella, "2025 DTC Digital Advertising Incrementality Benchmarks" · Haus (2025) · Enalitica, decision-maker trust in MMM vs. incrementality testing (2026) · Northbeam, Triple Whale (Moby), Recast, Measured, Haus - current DTC measurement tooling (2026)
Natalia Marianchyk
Marketing Specialist at NM Insight - a Berlin-based marketing professional focused on the infrastructure between acquisition and revenue.