SKAN vs Probabilistic vs SSOT: Finding Your Attribution Happy Place

If you’ve been in mobile marketing long enough, you’ve probably had that “what even is the truth?” moment staring at attribution reports.

Different numbers from different sources. Privacy changes yanking away your favorite levers. That roller-coaster ride where you’re guessing more than you’d like to admit.

In 2021, users started opting out of sharing their unique mobile identifier (known as IDFA). That’s how Apple’s App Tracking Transparency (ATT) changed everything.

One little pop-up, one little “Ask App Not to Track” and suddenly user-level IDs vanished. Until Q2 2025, only about 35% of users opted in (up slightly from 34.5% in Q2 2024 and 34% in Q2 2023). This means that the vast majority no longer share their data with companies. Deterministic attribution went out of the window.

Enter SKAdNetwork (SKAN), Apple’s privacy-first attribution framework. Not perfect. Not loved. But here to stay. And with Google’s Privacy Sandbox on the way, it’s not just an Apple thing.

Add in probabilistic models and MMP-led Single Source of Truth (SSOT), and marketers now juggle three attribution paths.

Here’s how they work, when to use them, and what’s coming next.

What SKAN Actually Is (and Isn’t)

Apple’s SKAdNetwork is attribution without device IDs. Instead, advertisers get anonymized postbacks from Apple with a delay.

Key features in SKAN 4.0+:

• No device IDs → privacy-first by design.
• Multiple postbacks → up to three, unlocking optimization signals beyond day 1.
• Conversion values & coarse values → track meaningful events.
• Hierarchical campaign IDs → more flexibility in structuring campaigns.
• Crowd anonymity tiers → the more installs you have, the more detail you see.

Why it matters for advertisers:

• Compliance-first → fully aligned with Apple’s rules.
• Cleaner data → SKAN installs aren’t marked as fraud.
• Future-proof → Apple’s roadmap is SKAN + AdAttributionKit, not device IDs.

But let’s be real: SKAN is slow to reveal performance (7–14 days), limited in granularity, and forces marketers to wait before optimizing.

Probabilistic: Fast but Fragile

Probabilistic attribution models user behavior (think device signals, IPs, timestamps) to estimate which ad drove an install.

Why marketers use it:

• Speed → early feedback within hours.
• Granularity → more data for daily bid and creative tweaks.

The catch:

• Data is modeled, not confirmed.
• Policies keep tightening (fingerprinting is explicitly banned by Apple).
• Works best at scale (5,000+ installs in Tier-1 markets).

If SKAN is “truth with a delay,” probabilistic is “fast but fuzzy.” Useful directionally, but not the hill to build a long-term strategy on.

SSOT: Reconciling SKAN and Modeling Data Into One View

Unified reporting views, also known as Single Source of Truth (SSOT), are built by MMPs, merging SKAN postbacks with probabilistic and modeled attribution to create one reconciled dataset.

Why marketers like it:

• Removes double-counting between SKAN and MMP data.
• Gives a cleaner, consistent report to share across teams.
• Helps avoid the dreaded “two dashboards, two answers” problem.

Limitations:

• Not its own attribution method, it’s a reconciliation layer.
• Still inherits the underlying trade-offs of SKAN and probabilistic.

Think of SSOT as the referee keeping your reporting consistent, not the player scoring goals.

How MMPs Handle SSOT

When SKAN came along, marketers suddenly had two sets of numbers: SKAN’s delayed postbacks and their MMP’s modeled reports. Cue confusion.

To fix this, MMPs created unified reporting tools. Basically, these are ways to merge SKAN with their own data and ‘deduplicate’ (or as they say, ‘dedupe’) it so you see one number.

Here’s how the big players do it:

Aarki’s Take

These tools are handy for reporting sanity. Nobody wants two dashboards telling different stories. But remember:

• SSOT is about reporting consistency, not a new attribution method.
• It still inherits SKAN’s delays and thresholds.
• The cleanest, most future-proof path is still running on SKAN itself.

SSOT can smooth the reporting bumps, but your optimization strategy still depends on whether you want clean data (SKAN), faster reads (probabilistic), or a merged view (SSOT).

How to Choose: A Practical Framework

Different advertisers need different setups. Here’s a simple way to think about it:

• Strict fraud rules? → Go SKAN-only. Fraudulent installs don’t get flagged because Apple controls the signal.
• Loose fraud rules, need volume? → Use SSOT (PB + SKAN hybrid) for scale.
• Low ATT opt-in? → SKAN is your default path to meaningful reach.
• Running remarketing with ATT-consented users? → Layer probabilistic retargeting on top.

What to Expect in a SKAN Campaign

Marketers often ask: “What happens in the first two weeks?”

• Day 1–5 → Only one postback, limited install data
• Day 7–14 → Delayed reports start trickling in. CPI and ROAS stabilize.
• Optimization begins → Conversion values (install → add-to-cart → purchase) kick in once privacy thresholds are met.

Pro tip: Don’t panic-pause a SKAN campaign on day 3. Give it the $500–$1,000/day runway it needs to cross thresholds and surface real performance.

Aarki’s Take: Balanced, Not Blinded

At Aarki, we don’t force a one-size-fits-all approach. We have:

• Full SKAN 4.0+ support → Multiple postbacks, coarse + fine values, campaign hierarchy.
• Flexible frameworks → SKAN-only, probabilistic, or SSOT, depending on your fraud rules and reporting goals.
• Proven integrations → AppsFlyer, Adjust, and Singular have already been validated.
• Enablement tools → Benchmarks, decision trees, fraud impact analysis to guide setup.

Aarki Is SKAN-Ready, With Privacy-First UA + RT at Scale, Now and Beyond

Most of iOS supply (95–98% today) already runs through SKAN. That means advertisers don’t lose scale when they choose compliance. The real challenge is in how campaigns are set up and optimized.

Aarki has rebuilt frameworks around SKAN 4.0+:

• Native postback ingestion that handles Apple’s delayed, anonymized data without hacks.
• Conversion value flexibility so schemas can track the right events across all three SKAN postbacks.
• Campaign hierarchy support to organize and optimize without hitting privacy thresholds too early.

Layer that with a balanced approach to user acquisition (UA) and retargeting (RT), both built to run under Apple’s privacy rules, and advertisers get scale without trading away compliance.

The goal is to give teams the choice: SKAN for clean reporting, SSOT when volume is the priority, and probabilistic remarketing where ATT consent allows.

Looking Ahead: SKAN 5.0 and Privacy Sandbox

Apple’s SKAN 5.0 promises:

• Faster postbacks (hours instead of days).
• Built-in incrementality (finally, a way to measure uplift).
• Retargeting support (privacy-safe remarketing).
  

On Android, Privacy Sandbox will roll out its own aggregated, privacy-first measurement. Translation: the privacy tide isn’t going out, it’s coming in stronger.

The Takeaway

• SKAN = clean, deterministic, privacy-compliant.
• Probabilistic = fast, directional, but policy-sensitive.
• SSOT = reconciled reporting, not its own method.

You don’t have to love SKAN. But you do have to live with it. And the earlier you start building your SKAN playbook, the easier the shift will be when SKAN 5.0 and Privacy Sandbox become the default reality.

Turn privacy rules aren’t going anywhere, but your growth can. Whether it’s SKAN, Probabilistic, or SSOT, all roads lead to growth with us. Claim your edge.