Past Sessions

• Event: MAU Vegas 25
• Date: Thursday, May 22, 2025
• Speakers: Eran Dunsky, Head of Product, AppsFlyer and Raye Askin, Global Product Lead for App Ads Measurement, Google
• Estimated read time: 6 to 8 minutes

Quick Read Summary

iOS attribution is no longer a reporting problem, it is a business confidence problem. Signal loss did not just reduce precision, it made budget decisions slower, harder to defend, and easier to get wrong.

The new reality is that deterministic attribution only covers a minority of iOS installs, even when opt in rates look healthy. When dual consent is required, usable IDFA availability drops to roughly 25 to 30 percent, which means most performance signals must be reconstructed, validated, and reconciled.

AppsFlyer and Google are pointing to the same solution pattern, combine deterministic signals where consent exists, probabilistic modeling where it does not, then validate, deduplicate, and report it through a single measurement layer you can actually use to run the business.

What follows is a practical explainer of the frameworks behind that pattern, and what mobile growth leaders should do next.

The real problem is fragmented truth, not missing data

Most teams describe the post ATT era as a loss of data. That is accurate, but incomplete. The more damaging shift is that teams lost a shared truth they could align around. When measurement becomes fragmented across SKAdNetwork style signals, self reporting network claims, and MMP level models, attribution is no longer a single answer. It becomes a negotiation between dashboards.

This is why the stakes are higher than reporting hygiene. Without a credible baseline, growth teams cannot reliably optimize creative, bids, or channel mix. Finance teams cannot trust payback curves. Product teams cannot separate acquisition quality from monetization changes. Measurement becomes “hard” not because people forgot how to do it, but because the system now produces multiple partial truths.

A useful way to frame the challenge is this, you are not trying to recover the past, you are trying to rebuild decision grade confidence under new constraints.

Why opt in rates can look fine, and attribution can still fail

One of the most misleading comfort signals in iOS measurement is a stable opt in rate. If opt in stabilizes around 50 percent, it is tempting to assume half your traffic is measurable in the old way. In practice, that is not what happens. The session highlighted the compounding effect of dual consent, once you account for it, IDFA availability falls closer to 25 to 30 percent.

That gap matters operationally:

• It pushes the majority of installs outside deterministic, IDFA based attribution.
• It increases reliance on modeled signals, which can drift, and are harder to compare across networks.
• It magnifies the cost of inconsistent definitions, especially around conversion windows, event schemas, and deduplication rules.

In other words, even if campaigns are performing, the organization may not be able to prove why, or replicate it, which directly impacts scaling decisions.

The winning pattern is hybrid attribution plus validation

The direction described in the session is not “choose probabilistic” or “choose deterministic.” It is to operate both at once, then validate what you can, and reconcile it into a single usable measurement layer.

AppsFlyer’s framing is explicit, Advanced SRN combines deterministic attribution for IDFA based traffic when dual consent exists, with probabilistic modeling for identifier less traffic, then feeds that into an attribution engine designed to produce more complete measurement.

Google’s framing is complementary, Integrated Conversion Measurement increases conversion observability and improves how conversions are represented in probabilistic reporting interfaces, particularly for iOS 14 plus devices, while maintaining privacy.

The strategic point is that hybrid is not a compromise. It is a deliberate architecture for a world where consent conditions vary by user, by region, and over time.

Stop treating modeled attribution as a backup plan. Treat it as a first class input, with governance. That means documented assumptions, consistent event definitions, and routine validation checks against business outcomes like revenue cohorts and retention.

Single Source of Truth is an operating system decision

Most leaders understand “single source of truth” as a reporting convenience. In the privacy era, it is closer to an operating system decision. If teams do not share a deduplicated view across SKAdNetwork style signals and MMP attribution, they do not share the same reality, and optimization becomes political.

AppsFlyer’s Single Source of Truth position is to bridge “scan based attribution” and AppsFlyer attribution, deduplicating signals across these methods so advertisers can make better decisions.

The specific outputs that matter are the ones finance and product care about:

• More conversions attributed, which can lower reported CPI and CPA when measurement gaps close.
• Regained cohort style metrics and down funnel event visibility, which are essential for LTV based optimization.
• More reliable LTV measurement “in a privacy centric way,” which is the real unlock for scaling iOS with confidence.

The session also pointed to a concrete outcome signal, iOS spend grew 77 percent year over year in Q1, attributed to recovered signals and improved measurement frameworks, although adoption remains uneven.

If your organization is still debating attribution source priority on a per channel basis, you are leaving performance on the table. Define your measurement hierarchy, decide how deduplication works, and align stakeholders on what “truth” means before you argue about optimization.

Self reporting networks were the missing piece, until they were not

A major measurement gap in recent years has been self reporting networks, which can provide conversion claims but often limit transparency and interoperability. The session positioned Advanced SRN as a way to “overcome signal loss” and support robust attribution and measurement across walled gardens, while maintaining privacy.

This matters because SKAdNetwork style signals alone often lack the granularity growth teams need for rapid iteration, and self reporting claims alone can be difficult to validate or compare. The core move is not to pick one, it is to combine them, then deduplicate.

AppsFlyer also referenced privacy preserving technologies such as data clean rooms and privacy enhancing technologies as part of the broader approach to building a richer, more accurate data picture.

Treat self reporting network measurement as an input that must be reconciled, not a truth that can be accepted at face value. Ask your partners, what is validated, what is modeled, what is deduplicated, and what is the error budget you should assume.

Google’s Integrated Conversion Measurement, device level innovation as measurement strategy

Google’s Integrated Conversion Measurement is positioned as a way to unlock more comprehensive, accurate attribution across iOS and Android campaigns, delivered through MMP interfaces like AppsFlyer, with real time, event level reporting, and without compromising privacy.

The key innovation described is an expanded on device conversion measurement approach. Earlier implementations leaned on first party data such as email addresses or phone numbers to improve observable conversions in iOS campaigns. The newer iteration extends to event level capability by using temporary, de identified app event data and cryptographic matching to improve measurement accuracy, with a strict privacy boundary, no user identifying information leaves the device or is disclosed externally, including to Google.

This is important for leaders because it shows where measurement is heading. The future is not bigger identifiers. It is privacy preserving computation, executed closer to the user, then surfaced as validated signals that platforms and MMPs can reconcile.

If your measurement stack is not keeping pace with SDK capabilities, you are choosing poorer data. Modern iOS attribution is increasingly gated by implementation readiness, not just media strategy.

How the collaboration works, and why the validation step is the point

The most operationally relevant part of the session is the flow between Google and AppsFlyer, because it shows how hybrid attribution becomes decision grade reporting.

At a high level:

• A user installs an app that has implemented the on device conversion measurement solution using event data.
• AppsFlyer tracks a first open and sends Google an attribution request for that install conversion.
• Google returns whether the conversion is deterministic or probabilistic, plus additional details such as timestamps and campaign IDs.
• AppsFlyer validates the response using proprietary models, and only validated claims move forward as candidates for last click attribution, then appear in Single Source of Truth reporting.

The validation step is the point, not a footnote. It is what converts cross network claims into something you can trust enough to optimize against, and defend internally.

When evaluating any “enhanced attribution” solution, ask where validation happens, what criteria are used, and how often models are recalibrated. A black box that cannot explain confidence is not a source of truth, it is just a different dashboard.

What to do next, a short adoption checklist for growth leaders

If you want to benefit from these frameworks, the work is not theoretical. It is operational, and it is measurable.

Focus on five moves:

• Audit your current iOS attribution coverage, quantify what percent of installs are deterministic, then what percent are modeled, and where gaps cluster by channel and geography.
• Align on a Single Source of Truth definition, including deduplication rules, reporting priorities, and the metrics that will govern budget decisions.
• Update implementation, the session called out adopting the latest Firebase SDK with on device event data capability, and updating AppsFlyer or your MMP SDK to support the latest solutions.
• Build a validation routine, compare reported conversion movement to business outcomes like revenue cohorts and retention, then flag divergence early.
• Re train optimization habits, shift from last click certainty to confidence based decisioning, where you incorporate probabilistic signals intentionally rather than reluctantly.

Conclusion, iOS attribution is being rebuilt, not restored

The post ATT era forced the industry to confront an uncomfortable truth, the old measurement model was fragile because it assumed plentiful identifiers. The new model is stronger because it assumes constraint, then designs around it.

The frameworks discussed here, hybrid attribution, privacy preserving measurement, validation, and a Single Source of Truth, are not vendor features. They are the architecture of modern mobile growth. Teams that adopt them early will move faster, spend with more confidence, and waste less time arguing about whose dashboard is correct.

The goal is not to recreate the past. It is to build iOS attribution that is resilient, privacy aligned, and credible enough to run the business.

Speakers

Eran Dunsky, Head of Product, AppsFlyer.  Eran leads product strategy at AppsFlyer, shaping the company’s privacy first measurement framework. He focuses on attribution innovation, including Advanced SRN and Single Source of Truth, with a mission to help marketers navigate the complexities of iOS attribution.

Raye Askin, Global Product Lead for App Ads Measurement, Google. Raye drives Google’s product direction for app ads measurement. She leads development of Integrated Conversion Measurement and advances Google’s on device conversion tracking systems, creating new ways for advertisers to measure performance securely and accurately.