SKAN 4 vs AdAttributionKit: How Apple Search Ads Attribution Works Now

If you searched for "SKAN 5.0" and landed here: it never shipped. Apple announced SKAdNetwork 5 at WWDC 2023, then quietly replaced it with a new framework, AdAttributionKit↗, introduced at WWDC 2024 and expanded at WWDC 2025. Most of what SKAN 5 promised — re-engagement attribution above all — arrived there instead.

That leaves iOS marketers in 2026 running two frameworks in parallel: SKAdNetwork 4.0, still the production standard for install attribution across the big networks, and AdAttributionKit, which extends the same on-device, aggregate model to re-engagement and alternative app marketplaces. If you run Apple Search Ads, you also have a third source of truth — Apple's own AdServices attribution — that neither of the other two replaces.

This guide covers what SKAN 4.0 actually does, what AdAttributionKit adds, who runs which version today, and how to plan measurement around the overlap.

What happened to SKAN 5.0

SKAN 5 was announced with four headline features: re-engagement attribution, faster postbacks (hours, not days), built-in incrementality measurement, and privacy-safe retargeting. None of it shipped under that name. Apple instead introduced AdAttributionKit as the umbrella framework, and re-engagement support landed there with iOS 18.4.

The practical consequence: any plan, vendor roadmap, or blog post that talks about "migrating to SKAN 5" is describing something that doesn't exist. Ask instead whether your MMP and your networks support AdAttributionKit, and to what extent.

From Limited to Layered: What SKAN 4.0 Changes

SKAN 4.0 introduces three major upgrades:

Multiple Conversion Windows

Instead of a single postback, SKAN 4.0 allows for up to three postbacks over time—covering 0-2 days, 3-7 days, and 8-35 days after the ad interaction. This longer conversion window is crucial for measuring events that don't happen immediately, such as subscriptions or in-app purchases (IAPs).

Crowd Anonymity Levels

Apple introduces privacy tiers (low, medium, high) based on how many installs your campaign gets. More installs = more data. At higher tiers, you receive more detailed postbacks. This incentivizes scale, making high-volume campaigns more valuable for insights.

Hierarchical Source IDs

Source identifiers now have a tiered structure—from 2 digits (low volume) to 4 digits (high volume)—allowing better segmentation of campaigns, ad groups, or even creatives. But again, the granularity you get depends on volume.

AdAttributionKit: what actually changed

AdAttributionKit keeps the parts of SKAN that make it privacy-preserving — on-device matching, aggregated postbacks, no user-level identifiers — and changes the scope:

• Re-engagement attribution. SKAN only ever attributed installs. AdAttributionKit can attribute a conversion after a user who already has the app taps an ad and opens it. This is the feature marketers waited three years for.
• Alternative app marketplaces. With third-party marketplaces on iOS in the EU, attribution had to work outside the App Store. AdAttributionKit is the framework that does.
• Interoperability with SKAN. Apple documents how AdAttributionKit and SKAdNetwork interoperate↗: an ad network on AdAttributionKit can still receive SKAN-style postbacks from apps that haven't adopted it, so the migration isn't a cliff.
• Same conversion-value model. Coarse and fine conversion values, multiple postback windows, and crowd-anonymity tiers carry over. If you designed a SKAN 4 conversion schema, it isn't wasted work.

What it does not change: postbacks are still delayed, still aggregated, and still gated by volume. AdAttributionKit is an extension of the privacy model, not a way around it.

Who runs which version in 2026

Adoption lags the framework by years, and it matters for what your reports mean. Industry reporting through 2026 puts most large ad platforms still primarily on SKAN 3 postbacks, with TikTok furthest along toward SKAN 4, and Meta, Google, and Snap largely on SKAN 3 with partial SKAN 4 support. AdAttributionKit adoption is earlier still, concentrated in networks that need re-engagement or EU marketplace coverage.

Two things follow:

1. Your SKAN 4 conversion schema is only as useful as the network reading it. A network on SKAN 3 collapses your three-window design into one postback. Check per network before you judge D7 revenue from postbacks.
2. Cross-network comparison on postback data alone is unreliable. Different versions, different windows, different anonymity thresholds. Anchor the comparison on something every network shares — realized revenue by install cohort from your subscription or MMP data — and treat postbacks as the per-network signal, not the scoreboard. We cover that reconciliation in how to calculate true ROAS from RevenueCat.

What This Means for Apple Search Ads

Apple Search Ads sits in an unusual position. It has its own deterministic attribution through the AdServices framework — a device-level token that returns campaign, ad group, and keyword IDs without an ATT prompt — and since 2025 it also emits SKAdNetwork and AdAttributionKit postbacks so MMPs can see it alongside every other network. The AdServices path is the one to build keyword-level ROAS on; the setup is in our RevenueCat + Apple Search Ads guide. SKAN 4.0 postbacks add a second, network-comparable view outside Search Ads' built-in reporting—especially when using third-party MMPs like Adjust↗ or AppsFlyer↗, or internal data pipelines.

Here's how it impacts your Apple Search Ads campaigns:

• Better ROAS Attribution for Delayed Conversions: With up to 35 days of tracking, subscription apps and IAP-heavy games can finally attribute late events more accurately. The math behind this is covered in detail in our piece on ROAS in IAP games.
• Increased Emphasis on Volume: To unlock granular postbacks and full Source ID range, campaigns need volume. This could push advertisers to consolidate budgets or optimize towards broader match types to scale up impressions and installs.
• Enhanced Creative & Keyword Testing: The hierarchical Source ID system makes it possible to track performance at a more granular level—if you hit the volume threshold. Smart keyword grouping and creative testing become essential.

Practical Example: A Gaming App Using Apple Search Ads

Let's say you're promoting a PvP mobile game with IAP monetization. Under SKAN 3.0, your visibility on purchases made after the first day was blurry at best. Now, with SKAN 4.0, you can track a user who installs on Day 0, plays casually for a few days, and then makes their first purchase on Day 5. That purchase can now be attributed properly—assuming your campaign volume is sufficient.

This changes how you model your return on ad spend (ROAS). No longer do you have to settle for early proxy metrics like retention or tutorial completion. You get actual revenue events, spaced across three postbacks, allowing for more accurate LTV modeling.

Challenges to Keep in Mind

Despite its improvements, SKAN 4.0 isn't perfect:

• Data Delay: Postbacks still come with delays (up to 24-48 hours), making real-time optimization difficult.
• Data Gaps for Low-Volume Campaigns: Smaller campaigns might still be stuck with limited visibility due to low crowd anonymity.
• Complex Implementation: Taking full advantage of SKAN 4.0 often requires working closely with MMPs, using Apple's API updates, and adjusting data infrastructure.

Best Practices for Advertisers

• Optimize for Early Signals + Late Events: Design user flows that encourage valuable actions early (within Day 0–2), but also track mid- to long-term conversion behavior.
• Consolidate Where It Counts: Focus your budget on campaigns that can scale, unlocking richer data from SKAN's tiered structure.
• Leverage ASA Insights + SKAN Data Together: Use Apple Search Ads' native reporting for short-term performance, and SKAN 4.0 data to model longer-term value.
• Keep Your MMP Updated: Work with your Mobile Measurement Partner to ensure full SKAN 4.0 integration. Many offer tools to map postbacks and model LTV effectively.

Final Thoughts

SKAN 4.0 set the measurement model iOS runs on, and AdAttributionKit extends it rather than replacing it. Nothing called SKAN 5 is coming to rescue anyone from postback delays or anonymity tiers. For Apple Search Ads, it unlocked long-awaited visibility into delayed conversions and deeper campaign segmentation.

But to benefit, advertisers must adapt their strategy: prioritize scale, update attribution models, and bridge first-party data with SKAN insights. Those who master this complexity will not only survive the post-IDFA era—they'll thrive in it. For broader UA economics in this new environment, see our analysis on whether paid marketing is getting too expensive.