Skip to content
AppDues
Log in
Contact salesSign up

SKAdNetwork for Subscription Apps: A Practical iOS Attribution Guide

Apple's privacy-first attribution changes what subscription apps can see about paid iOS users. This guide explains how SKAdNetwork operates, where its limits bite for subscription businesses, and how to build a workable measurement routine when user-level data is unavailable.

Why SKAdNetwork matters for subscription apps

Subscription apps depend on understanding which campaigns bring in users who eventually pay. To see why SKAdNetwork matters, it helps to start with the era before it.

Before Apple's App Tracking Transparency (ATT) framework and iOS 14, mobile marketers often describe the period as a golden age. User-level data was widely available, so installs, conversions, and engagement could be tied to specific users through the IDFA. Deterministic cohorts were practical and reliable, even though probabilistic matching also existed. MMPs held a near-monopoly on postback control and served as the single source of truth across self-attributed and non-self-attributed networks. ROAS and ROI were the main KPIs for campaign optimization and reporting. That comfortable setup let many marketers drift from the underlying business logic.

Even then, privacy pressure was building. Limited Ad Tracking (LAT) let iOS users avoid sharing their IDFA, and adoption grew substantially in 2019 and 2020. Privacy concerns were global: the EU's GDPR and California's CCPA signaled stricter data laws. Apple positioned itself as a privacy champion, with Tim Cook calling privacy a fundamental human right. 2020 marked a 180-degree turn for the status quo.

ATT became a reality in April 2021 with iOS 14.5. SKAdNetwork adoption was pushed worldwide, and ad networks and publishers had to implement the new system from scratch. The most consequential SKAN version to date, 3.0, introduced postbacks, conversion values, anonymity thresholds, and mapping. The high cost of privacy hit day-to-day operations: campaign quantity was limited by the postback digits available to ad networks; in-app events needed a new mapping approach; advertisers had little visibility into in-app activity; deterministic data at the creative level was lost; testing budgets shifted toward Android; and teams had to keep learning how to clear the anonymity thresholds. These challenges persist, partly because the system is technically intense and the gap between marketers and developers leaves few trustworthy sources to learn from.

The goal is not to recreate user-level attribution. It is to build a reliable proxy that tells you whether acquisition spend is producing more revenue than it costs.

How SKAdNetwork works in practice

Apple built SKAdNetwork as a consequence of ATT. It gives app developers (advertisers) a deterministic measurement tool that still protects user-level data. The ATT double-consent logic determines when a user can be prompted, but SKAN is the only attribution method that works independently of ATT status: even if ATT is not implemented in your app, SKAN can still operate and send postbacks. ATT prompting is not always possible. It is restricted for users under a certain age, for users who enabled LAT at the device level before iOS 14, and when the user has turned off the device-level "Allow Apps to Request to Track" master switch.

Apple's privacy-centric API has evolved over the years to version 4.0. Its core functionality stays the same, with minor improvements relative to its importance to app growth. The rest of this explanation focuses on 4.0; AdAttributionKit (AAK) is covered later.

At its core, SKAdNetwork lets an ad network sign an ad interaction, and Apple later sends a postback to that network and to the advertiser's configured endpoint. The advertiser's app updates a conversion value when the user triggers a chosen condition, such as a registration or purchase. A timer mechanism, controlled by Apple with a random element, determines when the winning postback is sent. Apple controls the countdown so advertisers cannot learn the exact install time.

Key terms to keep straight:

  • Advertised app: the app running ads to acquire users.
  • Publisher app: the app where the ad is shown.
  • Ad network: the company signing the ads and following Apple's rules.
  • Crowd anonymity tier: a framework that allocates users based on how hard they are to track deterministically. If you acquire ten users per day, the cohort will likely be signed with the lowest tier to protect each user's data. If you acquire 5,000 users per day, Apple will provide the highest tier and enable more granular data because it is harder to identify each user.
  • Conversion value: the in-app event signal a developer chooses to measure through SKAN.
  • Conditions: the triggers that update a conversion value, such as event counts, sessions, or total revenue.
  • Endpoint/info.plist: the destination where Apple sends a raw postback copy to the developer.

A useful example is a fitness subscription app running App Install campaigns through Meta. Meta registers under the SKAN program with an ad network ID. For Facebook the identifier is v9wttpbfk9.skadnetwork; for Instagram it is n38lu8286q.skadnetwork. When a user sees or clicks an ad and installs the app, SKAN recognizes the ad signature, starts the timer, and updates the conversion value as the user triggers conditions. After no more actions are done for 24 hours or the window closes, the highest conversion value tends to be sent to SKAN with a random delay. Apple sends the winning postback to Meta and a copy to the developer's endpoint. Second and third postbacks may apply depending on the crowd anonymity tier.

Three observations help explain SKAN's effect on the industry. First, the share of conversions carrying a media source fell, in part because of SKAN constraints such as anonymity thresholds and conversion windows. Second, retargeting conversions on iOS nearly vanished. Third, Apple Search Ads (ASA) emerged as one of the biggest winners over the last years, since it can measure deterministically regardless of ATT status.

What SKAdNetwork cannot do for subscription apps

A common misconception is that SKAN reports nothing when anonymity thresholds are triggered. In fact, SKAN still reports activity, but the conversion value parameter in the payload is set to null. That distinction matters: null postbacks are not missing data; they are data with the useful signal stripped out.

The practical limitations for subscription apps include:

  • Anonymity thresholds can prevent efficient business decisions, especially for smaller advertisers.
  • Conversion value updates depend on client-side user activity, so app retention and postback volume are closely correlated.
  • The payload does not contain ad-level creative data, which pushes advertisers to rely on ad networks for creative insight and often to use Android as a testing environment.
  • Server-to-server events require specific infrastructure to update the conversion value, particularly for trial-converted events.
  • There is a trade-off between event signals and event prioritization. With limited budget, optimizing for trial starts may generate more signals, while optimizing for subscription volume may be more valuable but slower.
  • Postback sequences cannot be connected across users, so replacing arrival date with an estimated install week can reduce discrepancies.
  • The fidelity type parameter is limited to click = 1 or view = 0, so a banner click and a one-minute video click look equally relevant.

On the positive side, SKAN works across ad networks (excluding Apple Search Ads), does not require ATT, is deterministic, and does not require an MMP or third-party vendor to operate.

Working with aggregated data: a practical checklist

Without a unique user ID, the goal is to build a consistent proxy rather than a perfect model. Use this checklist to keep the measurement honest:

  1. Do not attempt multi-touch attribution from SKAN postbacks alone. Randomized data and the absence of a stable ID make reliable MTA impossible.
  2. Treat null postbacks as real activity. Depending on the SKAN postback version, controlling the percentage of null values is critical because of crowd anonymity tiers or thresholds. Extrapolate null volume using non-attributed data as a proxy. For example, if you aim to calculate ROAS, multiply the number of null postbacks by the estimated ARPU Dx (x = day number). This way, you are counting everything.
  3. Use SKAN data as a free, cross-network proxy even if your company relies on probabilistic modeling elsewhere.
  4. Do not expect retargeting to work well with SKAN. Existing users are better reached through ads than through SKAN-based retargeting.
  5. Invest in conversion value mapping. It is the main lever for defining how useful the received data will be, though it still does not restore user-level data.
  6. Keep Android as a deterministic testing environment where possible, and use iOS SKAN data as a directional proxy.

When building a revenue model, establish a baseline and context first. The output is an estimate, so the UA team must accept a percentage of error. Ad network reporting is a proxy, not a source of truth. The accuracy of the revenue bucket depends on the conversion value mapping logic. Null values are essential to a realistic number because they represent revenue not tracked by SKAN. Self-reporting networks and non-self-reporting networks should be treated differently because non-SRNs are exposed to more ad fraud and fake numbers.

Revenue modeling notes for SKAN 4.0

If the app operates on SKAN 4.0, more postbacks may be passed back, which can increase model complexity. A P2 estimated revenue D35 metric with a multiplier, or a P3 estimated revenue D35 metric without a multiplier, may be appropriate. The model can also be adapted to each ad network's status: for example, Meta might rely on a P1 approach only, while TikTok and Google could add P2 and P3 metrics.

The metric is called estimated because the Px revenue value is imported from a median value within the conversion value mapping revenue range. Granularity depends on the postback source. If enriched postbacks are used, campaign name and country may be available for breakdowns; otherwise, data is hosted only at the ad network level.

If Px equals Dx, meaning P1 = D2, P2 = D7, or P3 = D35, the formula should not include a multiplier because the Px revenue is already the absolute total. The P1, P2, and P3 estimated revenue metrics correspond to those day windows, and the P2 metric includes a multiplier while the P3 metric does not.

Subscription-specific risks to manage

Free trials are a significant problem when building a ROAS system. The origin of the issue lies in the SKAN 3.0 conversion window; SKAN 4.0 changed some of this, but caveats remain. When apps scale pricing experiments, teams often forget about the revenue buckets created in the mapping, which can distort revenue reporting. Because SKAN depends on user activity to update conversion values, measuring events like trial_converted requires a specific server-to-server implementation. Down-funnel measurement therefore depends on app retention.

How the ad network landscape shifted

The picture is contested. As aggregated measurement (AEM) gains ground, and as non-SAN networks such as AppLovin try to step away from Apple's deterministic-centric API, the verdict on who gained and who lost depends on whom you ask. Still, the source's dated assessment names some clear winners and losers over the years.

On the winning side it places Apple Search Ads, web-to-app, TikTok, and influencers. On the losing side it places Google, Meta, X, and Snapchat. It also argues that Apple itself has been among the biggest losers over time, for four reasons: many ad networks still lean on IP-based matching and device IDs, to the point where AEM matters to many advertisers; non-SAN networks such as AppLovin or Unity have not adopted the technology, so gaming apps keep relying on fingerprinting; releasing a major new system without community support left app developers struggling to understand it; and Apple pushed apps into a binary choice between paying for an MMP or going without.

Treat these winner/loser judgments as dated captured opinion rather than a current market ranking.

AdAttributionKit: what changes and what does not

AdAttributionKit (AAK) is built on top of SKAN and adds tools, but not as many as the industry hoped. The most important feature is re-engagement, which allows advertisers to track conversions from ads clicked by users who already installed the app. This could matter for apps that drive retention and upsell existing customers. AAK also supports alternative marketplaces, adds fraud-prevention changes such as requiring ads to be displayed in the foreground and limiting timer abuse, and includes a developer mode for easier testing.

SKAN and AAK will operate in parallel, and ad networks already registered on the SKAN program will not need to do anything to start working with AAK. The real game changer over the following years is likely to be retargeting rather than measurement granularity. Apple's improvements focused on infrastructure topics related to testing and ad fraud when the industry's pain was around measurement capabilities.

Adoption is expected to be slow. Some months after iOS 18 adoption, time will tell whether Apple does something else to force wide adoption. If nothing changes in the short term, it is risky to say that deterministic attribution will prevail over fingerprinting. Privacy remains the center of the discussion, and developers need more resources to use these tools profitably. Whatever the framework, the primary KPI remains the same: more money coming in than going out.

Frequently asked questions

What is SKAdNetwork in simple terms?

SKAdNetwork is Apple's privacy-preserving attribution API. An ad network signs an ad interaction, the advertised app updates a conversion value when the user triggers chosen conditions, and Apple later sends a postback to the ad network and to the developer's endpoint. It works independently of ATT status.

Does SKAdNetwork require App Tracking Transparency?

No. SKAN operates independently of ATT status. Even if ATT is not implemented in your app, SKAN can still run and send postbacks. ATT is a separate consent framework, and some users cannot be prompted for ATT: users under a certain age, users who enabled LAT at the device level before iOS 14, and users who have turned off the device-level 'Allow Apps to Request to Track' master switch.

Does SKAdNetwork stop reporting when anonymity thresholds are triggered?

No. SKAN still reports the activity, but the conversion value parameter in the payload is set to null. Null postbacks represent real activity whose useful signal has been stripped out, so they should be counted and extrapolated rather than ignored.

How is AdAttributionKit different from SKAdNetwork?

AAK is built on top of SKAN. Its most important addition is re-engagement, which lets advertisers track conversions from ads clicked by users who already installed the app. It also supports alternative marketplaces, adds fraud-prevention requirements, and includes a developer mode for testing. SKAN and AAK operate in parallel, and ad networks already registered on SKAN do not need to do anything to start working with AAK.

Do I need an MMP to use SKAdNetwork?

No. You can work with SKAN using the ad network SDK or by writing code to create the conversion value mapping. No third-party company or vendor is required to use it.