iOS Attribution After ATT: Understanding SKAdNetwork 4

Key Takeaways

• App Tracking Transparency ended deterministic user-level attribution on iOS, because most users decline cross-app tracking, so SKAdNetwork became the primary privacy-preserving measurement framework.
• SKAdNetwork attribution is done by Apple, not the advertiser, and returns aggregated, delayed postbacks with a limited conversion value — not user-level event data.
• SKAdNetwork 4 expanded the framework: multiple postbacks over a longer window, coarse and fine conversion values, hierarchical source identifiers for more granularity, and web-to-app attribution.
• The conversion-value schema — how you encode what a user did into the limited bits available — is the most important design decision, because it determines what you can actually learn and optimise toward.
• Measurement gaps remain by design: aggregation, delay, thresholds that suppress low-volume data, and no user-level detail — so iOS measurement is inherently coarser than the deterministic era.
• Profitable iOS acquisition means designing a good conversion-value schema, using aggregated signal well, and complementing it with modelling — not pretending deterministic attribution still exists.

What App Tracking Transparency Changed

App Tracking Transparency, or ATT, is the Apple policy that requires apps to explicitly ask a user's permission before tracking them across other companies' apps and websites — the well-known prompt asking whether an app may track you — and its effect on advertising measurement was seismic because most users, when asked, decline. Before ATT, iOS advertising relied on a device identifier (the IDFA) that let advertisers deterministically connect an ad impression or click to a later app install and in-app behaviour at the individual level, exactly as third-party cookies did on the web. ATT did not delete that identifier, but it made access to it contingent on user permission that most users refuse, which for practical purposes removed deterministic user-level tracking from the majority of iOS traffic. The identifier still exists; you just cannot use it for most users, which amounts to the same thing.

The consequence is that the deterministic, user-level attribution that iOS advertisers depended on — knowing that this specific user saw this specific ad and then installed and did these specific things — is gone for most of the audience, and no amount of wishing brings it back. This is the crucial mindset shift: iOS measurement after ATT is fundamentally different, not just degraded, and advertisers who try to recreate deterministic per-user attribution through workarounds are both fighting Apple's privacy direction (which risks their apps) and chasing a capability that is structurally unavailable. The correct response is to work within the privacy-preserving framework Apple provides, which is SKAdNetwork, and to build a measurement approach that produces good decisions from aggregated, privacy-first signal rather than one that depends on the user-level detail that no longer exists.

It is worth being clear about why this happened and why it is permanent, because that clarity is what stops advertisers from waiting for it to reverse. ATT is part of a broader, industry-wide, regulator-backed shift away from cross-party individual tracking, the same shift driving third-party cookie deprecation on the web, and it reflects a durable change in what users, regulators and platforms will accept. It is not a temporary restriction that will loosen; it is the new baseline, and Apple has been extending privacy protections, not relaxing them. So the advertisers who thrive on iOS are the ones who accepted early that deterministic user-level attribution is over and invested in doing privacy-preserving measurement well, while those still trying to reconstruct the old world spend their energy fighting a tide that is not turning. The rest of this guide is how to do the new measurement well.

How SKAdNetwork Works

SKAdNetwork is Apple's framework for attributing app installs (and, increasingly, post-install events) to ad campaigns without exposing user-level data to advertisers, and its defining characteristic is that Apple, not the advertiser or the ad network, performs the attribution. When a user sees or clicks an ad and later installs and uses the app, Apple's system determines that the install is attributable to that ad campaign and sends a 'postback' — a notification of the attributed install and some limited information about what the user did — but it does so in a way designed to prevent the advertiser from connecting the data to an individual. The advertiser receives aggregate signal about which campaigns drove installs and roughly what those users did, without receiving the user-level detail that would allow individual tracking. This is a fundamentally different model from the old one: the platform mediates and aggregates the attribution rather than handing the advertiser the raw connection between ad and user.

The key mechanism advertisers must understand is the conversion value. Because Apple will not send user-level event data, SKAdNetwork provides a small, limited 'conversion value' — a constrained amount of information — that the app can set to encode what the user did after installing (for example, whether they registered, made a purchase, reached a certain value tier). This conversion value is the only window advertisers have into post-install behaviour through SKAdNetwork, and it is deliberately limited in size, so you cannot pass rich event detail; you have to decide how to encode the most decision-useful information about user behaviour into the small space available. The postback then returns this conversion value in aggregate, telling you not just that a campaign drove installs but roughly what those users did, within the limits of what you chose to encode.

Two further design features shape everything about SKAdNetwork measurement: delay and thresholds. The postbacks are deliberately delayed — Apple introduces timing randomisation so that the attribution cannot be tied to a specific user by its timing — which means SKAdNetwork data arrives later than the real-time measurement advertisers were used to, complicating rapid optimisation. And Apple applies privacy thresholds: when the volume of installs or events for a given campaign is too low, data may be suppressed or coarsened to prevent it from being used to identify individuals, so low-volume campaigns and granular breakdowns can return little or no usable signal. These features — Apple-mediated attribution, a limited conversion value, deliberate delay, and privacy thresholds — are not bugs to be worked around but the intentional design of a privacy-preserving system, and understanding them is what lets you design measurement that works within them rather than fighting them.

What SKAdNetwork 4 Added

SKAdNetwork 4 was a significant expansion of the framework that made privacy-preserving iOS measurement meaningfully more useful, while keeping the fundamentally aggregated, delayed, privacy-first design intact. The headline additions addressed several of the sharpest limitations of earlier versions, giving advertisers more signal to work with without reintroducing user-level tracking. The most important is multiple postbacks: where earlier SKAdNetwork sent a single postback shortly after install, SKAdNetwork 4 sends multiple postbacks over a longer window after install, which lets advertisers see not just immediate post-install behaviour but how users develop over time — crucial for measuring the delayed value that matters for apps where the important behaviour (subscription, repeat purchase, retention) happens days after install, not minutes.

SKAdNetwork 4 also introduced two tiers of conversion value: a fine conversion value (the richer, more granular value available when volume is high enough to protect privacy) and a coarse conversion value (a lower-granularity value — low, medium, high — available even at lower volumes where the fine value would be suppressed). This tiered approach directly addresses the threshold problem: instead of getting no signal when volume is low, advertisers get at least a coarse signal, so more campaigns return usable data. The framework also added hierarchical source identifiers, which expand the campaign identifier space and allow more granular attribution (more campaign, and in some conditions ad-level, breakdown) than the very limited campaign IDs of earlier versions, so advertisers can attribute to a more granular level of their campaign structure when volume permits.

Finally, SKAdNetwork 4 added support for web-to-app attribution, allowing installs driven by ads on the web (not just within apps) to be attributed through the framework, which closed an important gap for advertisers whose iOS acquisition includes web ad placements. Taken together, these additions — multiple postbacks over a longer window, coarse and fine conversion values, hierarchical source identifiers, and web-to-app attribution — make SKAdNetwork 4 substantially more capable than its predecessors, letting advertisers measure delayed value, get signal even at lower volumes, attribute more granularly, and cover web-driven installs, all while preserving the privacy-first design. The upshot is that privacy-preserving iOS measurement is far more workable than it was in the immediate aftermath of ATT, provided advertisers understand and design around the framework's structure rather than lamenting the loss of the deterministic era.

Designing Your Conversion-Value Schema

Because the conversion value is the primary window into what users do after installing, designing your conversion-value schema — deciding how to encode user behaviour into the limited space SKAdNetwork provides — is the single most important measurement decision an iOS advertiser makes, and it determines what you can actually learn and optimise toward. The constraint is that the space is small, so you cannot encode everything; you have to choose which aspects of post-install behaviour are most decision-useful and encode those, trading off granularity across the dimensions you care about. This is a genuine design problem: a schema optimised to measure early registration tells you nothing about purchase value, while one optimised for purchase value tells you nothing about the engagement steps in between, so you have to align the schema with the behaviour that actually matters for your business and your optimisation.

The right schema depends on your app's economics and what you need to optimise toward. If your key outcome is a purchase whose value varies, you might encode value tiers so you can distinguish high-value from low-value installs and optimise toward the former; if your key outcome is reaching an engagement milestone that predicts retention, you might encode the funnel steps toward that milestone; if delayed value matters (with SKAdNetwork 4's multiple postbacks), you design the schema to capture how value develops over the postback windows. The discipline is to work backward from the decisions you need the data to inform — which campaigns are acquiring valuable users, which creative drives valuable installs, where to allocate budget — and design the schema to encode exactly the signal those decisions require, within the space available, rather than trying to capture everything and capturing nothing usefully.

Getting the schema right is where much of the skill of post-ATT iOS measurement lives, and it is easy to get wrong by either over-reaching (trying to encode too much and ending up with signal too diluted to act on) or under-reaching (encoding too little and being unable to distinguish valuable from worthless installs). A well-designed schema, aligned with your economics and your optimisation decisions, turns SKAdNetwork's limited conversion value into a genuinely useful signal about which campaigns and creative acquire the users you want; a poorly-designed one turns it into noise. And because SKAdNetwork 4 added coarse and fine values and multiple postbacks, the schema design now includes deciding how to use those — what to encode in the coarse value for low-volume situations, how to structure the value across the multiple postback windows — which makes good schema design both more powerful and more involved than it was. This is the foundation on which all your iOS optimisation rests, so it deserves real thought.

The Measurement Gaps That Remain

Even with SKAdNetwork 4's improvements, real measurement gaps remain by design, and pretending they do not exist leads to bad decisions, so profitable iOS advertising depends on understanding exactly what you cannot see and planning around it. The first gap is aggregation: you get campaign-level (and with SKAdNetwork 4, somewhat more granular) aggregate signal, not user-level detail, so you cannot analyse individual user journeys, build user-level models from the SKAdNetwork data, or do the granular cohort analysis the deterministic era allowed. The second gap is delay: postbacks arrive later than real-time data, and are timing-randomised, so you cannot optimise campaigns as rapidly or tie behaviour to precise timing, which slows the feedback loop and requires more patience in reading results.

The third gap is thresholds and suppression: Apple's privacy protections mean that low-volume campaigns, granular breakdowns, and rare events may return coarsened or no data, so smaller campaigns and long-tail analysis are systematically under-measured, and you can be flying partly blind on exactly the granular questions you most want answered. The fourth is the limited conversion value itself: no matter how well you design the schema, you can only encode a small amount of information, so there is post-install behaviour you simply cannot see through SKAdNetwork, and rich behavioural analysis of your acquired users has to come from your own first-party in-app data, not from the attribution framework. These gaps are not failures of the framework; they are the intended trade-offs of a privacy-preserving system, and they define the boundaries of what SKAdNetwork can tell you.

The correct response to these gaps is not to try to circumvent them — which fights Apple's privacy direction and risks your app — but to complement SKAdNetwork with other measurement that fills the gaps in privacy-preserving ways. Your own first-party in-app data gives you the rich, user-level understanding of your acquired users that SKAdNetwork cannot (once they are your users, you can measure their behaviour in your app, subject to your own privacy practices). Aggregated modelling — media-mix modelling and incrementality testing — gives you the cross-channel and causal picture that user-level attribution used to provide, using aggregate data that does not depend on individual tracking. And SKAdNetwork itself, well-configured with a good conversion-value schema, gives you the campaign-level acquisition signal. Together, these three — SKAdNetwork for privacy-preserving acquisition attribution, first-party in-app data for user understanding, and modelling for the cross-channel causal picture — constitute a complete iOS measurement approach that works within the privacy-first world rather than fighting it.

Running Profitable iOS Acquisition Now

Putting it together, running profitable iOS acquisition after ATT means accepting the privacy-preserving framework and getting genuinely good at working within it, rather than spending energy mourning or circumventing the deterministic era. The foundation is a well-designed conversion-value schema aligned with your economics, so that the signal you do get from SKAdNetwork is as decision-useful as possible — this is the highest-leverage measurement investment, because it determines the quality of every campaign decision downstream. On top of that, you use SKAdNetwork 4's capabilities well: reading the multiple postbacks to understand delayed value, using coarse values to keep signal flowing at lower volumes, and attributing at the granularity the hierarchical source identifiers and your volume allow, so you extract the maximum usable signal the framework provides.

Because SKAdNetwork signal is aggregated and delayed, you also adapt your optimisation approach to match: less rapid, granular, real-time tweaking (which the data no longer supports well) and more structural, patient, aggregate-informed decision-making, giving campaigns time to accumulate enough signal to read, structuring campaigns so that volume concentrates enough to clear privacy thresholds, and making budget and creative decisions on the aggregate signal rather than chasing user-level detail that is not there. This is a genuine shift in operating style — from the fast, granular optimisation the deterministic era allowed to a more measured, structural approach suited to aggregated privacy-preserving data — and advertisers who adapt their operating rhythm to the data they actually have make better decisions than those trying to run their old playbook on data that no longer supports it.

Finally, you complement SKAdNetwork with the first-party and modelling layers, so your total picture is complete even though no single source is. Your first-party in-app data tells you what your acquired users are actually worth and how they behave; your modelling (MMM and incrementality) tells you what your marketing is actually causing across channels; and SKAdNetwork tells you which campaigns are driving installs at what apparent value. A brand that runs all three in concert has a robust, privacy-compliant iOS measurement approach that supports profitable acquisition, while a brand relying on SKAdNetwork alone (and frustrated by its gaps) or trying to reconstruct deterministic tracking (and fighting Apple) does not. The winners on iOS after ATT are not the ones who found a clever workaround; they are the ones who designed a good conversion-value schema, used the framework's real capabilities well, adapted their operating style to aggregated signal, and complemented it with first-party data and modelling — which is exactly the durable, privacy-first measurement approach the whole industry is converging on.

Methodology & Fairness

A note on how to read this. This is an educational guide published by Fluxsy, a performance marketing partner, so weigh our perspective accordingly. Platform mechanics and privacy rules change frequently; verify the specifics described here against the current official documentation before you implement. Where we name tools, platforms or companies we describe them by their genuine public positioning, not as endorsements. We have avoided inventing statistics, benchmarks or results — the durable value here is the framework and the reasoning, which hold even as the specific implementation details move. Measure against your own data before concluding, because your results depend on your stack, your market and your configuration.