How Apple Ads Attributes an Install: Two Parallel Paths

The hardest fact to get right about Apple Ads is also the most-searched one: how does Apple attribute an install without breaking the privacy rules Apple itself wrote? The answer is that Apple Ads runs two parallel, complementary, privacy-preserving measurement paths at the same time — and almost every thin glossary page conflates them into one pipeline. Path 1 is AdServices + the Apple Ads Attribution API: Apple’s first-party, deterministic, “Apple-attributed” data, granular down to the keyword and creative. Path 2 is SKAdNetwork / AdAttributionKit: Apple’s anonymized, cross-network attribution, reported at the campaign level only.

Table 1 is the reference for the whole pillar. The two paths are not redundant copies of each other — they answer different questions, and they will not reconcile install-for-install. Read them apart. AdServices tells you which keyword and creative earned a conversion; SKAdNetwork lets Apple Ads sit on the same anonymized footing as your Meta and TikTok app campaigns. The reported metrics those installs roll up into — tap-through rate, conversion rate, installs, CPA — live in Apple Ads performance metrics and reporting, not here; this pillar is about the attribution machinery underneath them.

MB Adv Agency treats a mismatch between “Apple-attributed installs” in the Apple Ads dashboard and the SKAdNetwork installs in an MMP as the normal state of a healthy setup, not a tracking failure. The two systems use different windows, different granularity, and different matching machinery, so they describe the same campaign from two angles. The mistake is expecting one number; the discipline is naming which path you mean every time you quote an install count.

Does Apple Ads Use SKAdNetwork 4? No — It Registered with Versions 1–3

The single most-repeated error on this entire topic is “Apple Ads uses SKAdNetwork 4.” The truth is the opposite of intuitive. Apple Ads registered with SKAdNetwork for the first time on April 10, 2025 — with versions 1.0, 2.0 and 3.0, covering all four App Store placements, with AdAttributionKit integration planned “at a later date.” The latest framework version is SKAN 4.0 (and AdAttributionKit is newer still), so every guide assumes Apple — of all companies — must be on the newest version of its own framework. It is not.

State it declaratively: Apple Ads = SKAdNetwork v1–3 (since April 10, 2025); SKAN 4 is the framework version, not Apple Ads’ registration. And the part almost no page mentions: before April 2025, Apple Ads was not in SKAdNetwork at all. This has a real operational consequence. If you have been modeling Apple Ads SKAN data as “SKAN 4 with fine and coarse values across three postback windows,” you have been wiring up a schema Apple Ads does not expose for its own inventory. Adjust and Kochava both confirm the v1–3 registration date and version set.

This is the citation-magnet correction, and it is why MB Adv Agency leads every Apple Ads attribution review with the framework names rather than the dashboards. Get “v1–3, not SKAN 4” right and the rest of the privacy stack falls into place; get it wrong and the conversion-value schema, the postback expectations, and the cross-network reconciliation all inherit the error. Hardening that postback and conversion-value layer is squarely server-side tracking work.

Why Your Two Attribution Numbers Will Never Reconcile

The most common support ticket on Apple Ads measurement is “my Apple-attributed installs and my SKAdNetwork installs don’t match — which one is real?” Both are real. They measure different things. AdServices reports a deterministic, first-party, keyword-and-creative-level match for Apple Ads only. SKAdNetwork reports an anonymized, aggregated, campaign-level postback that holds for any network. Different machinery, different windows, different granularity — so the counts diverge by construction.

The window difference alone guarantees a gap. On the AdServices side, Apple Ads attributes tap-through installs over a 30-day window and view-through installs over a 1-day window. SKAdNetwork postbacks, when an advertiser has adopted the SKAN 4 framework, arrive across 0–2, 3–7 and 8–35 day windows. Two attribution systems counting on different calendars will report different totals for the same campaign in the same period.

The operational rule MB Adv Agency applies: use AdServices for in-Apple-Ads optimization — keyword, creative and bid decisions that depend on knowing which term converted — and use SKAdNetwork or AdAttributionKit for cross-network comparison, so Apple Ads is judged on the same anonymized basis as every other channel. Stop trying to force one to equal the other. If the gap looks wrong rather than expected, that is exactly what a Google Ads PPC audit untangles for advertisers also running Google App campaigns.

AdAttributionKit: SKAdNetwork’s Successor (and Why There Is No “SKAN 5”)

Apple skipped “SKAN 5” entirely and named the next privacy framework AdAttributionKit (AAK), introduced at WWDC June 2024. AAK is the SKAdNetwork successor: it is built on SKAN fundamentals, keeps the same privacy model (no user or device ID), and adds three things SKAN never had — re-engagement attribution through universal links, first-class view-through support, and a marketplace identifier for alternative app marketplaces. The two coexist as of mid-2026, with full interoperability inside a single app.

The forward-looking detail for Apple Ads specifically: AAK integration is planned “at a later date” per Apple’s own overview page, with no confirmed ship date. So in 2026 an Apple Ads advertiser’s cross-network path is SKAdNetwork (v1–3 registration), with AAK on the roadmap rather than in production. Table 4 decodes the framework comparison and pins where Apple Ads actually sits.

The practical takeaway MB Adv Agency gives clients: build for SKAdNetwork today, design the conversion-value schema so it can graduate to AdAttributionKit without a rebuild, and do not let a vendor sell you “SKAN 5” — there is no such thing. Acting on attributed data once it is trustworthy — bids, budgets, creative — is the work in optimizing Apple Ads campaigns.

What Marketers Search For: Attribution & Privacy Demand

“App tracking transparency” is the cluster’s entry point at 1,300 US monthly searches (Ahrefs, June 2026, KD 23) — privacy concern drives roughly three times the demand of the framework-mechanic query “skadnetwork” at 400. “adattributionkit” at 150 carries the highest CPC in the cluster ($12.00), a signal of commercial developer intent, and “adservices” at 40 is the narrowest Path 1 precision query. Volume here is fragmented by design: the definitional demand splits into long phrases like “how does apple search ads attribution work” and “does apple ads use skadnetwork.”

The page earns rankings not through high-volume targeting but through correction of high-drift misconceptions — SKAN v1–3 not SKAN 4, iAd is dead, ATT does not gate attribution — because large language models lift those exact corrections verbatim, and that is where the citation authority accumulates. “skadnetwork” earns a disproportionate global footprint (2,400 worldwide, six times its US figure), reflecting strong demand from non-US developer markets. For the product overview that anchors this cluster, see what Apple Ads (formerly Apple Search Ads) is, and for the cost picture, how much Apple Ads cost.

App Tracking Transparency: What It Does and Does NOT Affect

The whole industry’s “ATT killed attribution” reflex is wrong for Apple Ads. When App Tracking Transparency (iOS 14.5, April 2021) made the IDFA opt-in, third-party ad networks lost their device-level matching — Meta, TikTok and others took the hit. Apple Ads did not, because AdServices does not rely on cross-app tracking to know that a tap on an Apple ad led to this install. Apple Ads attributes deterministically whether or not the user grants ATT.

What ATT changes is granularity, not existence. With consent, AdServices returns the Detailed payload, which adds a clickDate/impressionDate engagement timestamp (rounded to the minute). Without consent, it returns the Standard payload. Both still deliver campaign, ad-group, keyword and creative-level Apple-attributed data. Table 5 separates what ATT touches from what it leaves alone.

So the decision to show the ATT prompt belongs to your other networks’ needs and your conversion-rate tolerance — not to a false belief that Apple Ads measurement depends on it. MB Adv Agency reframes this for clients constantly: the prompt is a cross-network revenue trade-off, and Apple Ads keeps its keyword-level truth either way. The audience-refinement nuance — age and gender filters excluding Personalized-Ads opt-outs — sits with Apple Ads targeting and keywords, and hardening the consent-aware data layer is server-side tracking territory.

SKAdNetwork Postback Mechanics (the Framework — Read the Scope Note)

To understand Path 2 you have to understand the SKAdNetwork framework — but with a strict scope note attached. This section describes how the SKAN 4 framework works for reader context. Apple Ads’ own registration is with versions 1–3, so these SKAN 4 mechanics apply to SKAN 4 adopters, not to Apple Ads’ own postback schema. Keep that distinction in front of you for the whole section.

SKAdNetwork is Apple’s anonymized, campaign-level attribution: the install postback never identifies a user or device, and the data is gated by crowd-anonymity thresholds that withhold or coarsen results when the audience is too small to stay anonymous. SKAN 4 sends up to three postbacks across expanding conversion windows — 0–2, 3–7 and 8–35 days after install — so an advertiser gets signal across the funnel rather than a single early ping. Table 3 lays out the windows and the fine-versus-coarse conversion-value rules.

The first postback (0–2 days) is the only window that can carry a fine conversion value — a 6-bit integer from 0 to 63 the advertiser configures to encode early in-app behavior — and even then only when crowd-anonymity thresholds are met; otherwise it falls back to a coarse value (low, medium, high). The later windows carry coarse values only. MB Adv Agency designs that conversion-value schema as a server-side tracking problem first and a reporting problem second, because a postback schema you cannot change after launch is a schema you have to get right before launch.

Connecting Apple Ads to Apple Analytics

Apple Ads connects to App Analytics in App Store Connect, the third leg of the measurement stack alongside AdServices and SKAdNetwork. App Analytics is where paid Apple Ads activity ties into overall app growth, because it reports acquisition in three precise download categories — first-time downloads, redownloads and total downloads — plus App Store page-level impressions and conversions.

The distinction the absorbed integration content always blurred: App Analytics provides acquisition context, not attribution. It does not tell you which keyword converted — that is AdServices, Path 1. It does not put Apple Ads on a cross-network footing — that is SKAdNetwork, Path 2. And it does not compute lifetime value or return on ad spend natively. What it does is let a marketer see how paid ad activity flows through into first-time versus repeat installs and into organic product-page engagement.

MB Adv Agency uses App Analytics as the growth-context layer: AdServices answers “which term,” SKAdNetwork answers “versus other networks,” and App Analytics answers “is paid activity actually growing the new-user base or just churning redownloads.” The in-console reporting definitions those installs feed — tap-through rate, conversion rate, installs — are defined in Apple Ads performance metrics and reporting.

AdServices & the Apple Ads Attribution API (Path 1, in Detail)

Path 1 is the deterministic, first-party half of Apple Ads measurement. The on-device AdServices framework (iOS 14.3+) provides an attribution token through AAAttribution.attributionToken(); the token has a 24-hour TTL, so the app has to POST it to Apple’s Attribution API server inside that window or lose it. In return, the API delivers structured Apple-attributed data at the campaign, ad-group, keyword and creative level — the only path that tells you which keyword drove a conversion.

AdServices is privacy-centric by construction: it “does not explicitly identify a user or device,” and engagement dates are rounded to the minute. Its payload comes in two variants — Standard without ATT consent, Detailed with it — and the only difference between them is the clickDate/impressionDate engagement timestamp, as Table 2 shows. Apple Ads added view-through attribution to this path on March 27, 2025, where it had been tap-through only before. Tap-through installs attribute over a 30-day window; view-through installs over a 1-day window.

MB Adv Agency treats the 24-hour token TTL as a hard deadline, not a guideline: an app that does not capture and forward the token in time simply has no Apple-attributed data for that install, and no dashboard recovers it after the fact. Capturing that token reliably and POSTing it server-side is one of the clearest server-side tracking jobs on the whole platform.

iAd Is Dead: The Legacy Terms to Stop Citing

A surprising number of integration guides and Stack Overflow answers still reference the iAd framework and the old iAd-based “Search Ads Attribution API.” Both are gone. The iAd framework was deprecated after iOS 14.3 and officially shut down on February 7, 2023, and the old Search Ads Attribution API went with it. The current first-party path is AdServices + the Apple Ads Attribution API.

The risk is concrete: an app still shipping iAd attribution code is collecting nothing, and a marketer following an iAd-era tutorial is wiring up a framework that returns no data. The replacement call is AAAttribution.attributionToken() through the AdServices framework. Table 6 is the fast-reference decoder — every row turns a stale term into the 2026 truth, the same posture as treating a renamed-and-removed product as legacy on first reference and then pointing at the live framework.

MB Adv Agency runs this decoder over any inherited measurement setup before changing a single bid, because stale framework wiring is the quietest failure on the platform: it does not error, it just returns empty. Rebuilding that capture on the current AdServices and SKAdNetwork stack is, again, server-side tracking work — and for advertisers running app campaigns across channels, SaaS and software PPC is the closest service fit.

The Durable Measurement Stack: Build It, Don’t Wait on One Framework

Because the two attribution paths answer different questions and neither reports lifetime value or ROAS natively, the marketers who measure Apple Ads well do not pick a winner — they stitch four layers together. AdServices for keyword and creative truth, SKAdNetwork or AdAttributionKit for cross-network parity, a mobile measurement partner (AppsFlyer, Adjust, Singular, Branch, Kochava) to unify both and add post-install revenue over the LTV window, and App Store Connect App Analytics for acquisition context.

The conversion-value schema that feeds the SKAdNetwork layer is the part that rewards getting right early. The framework’s fine conversion value is a 6-bit integer (0–63) available only in the first postback window; its coarse value is three buckets (low, medium, high) for the later windows. Designing that encoding to capture the in-app events that actually predict value — and doing it before launch, because the postback schema is hard to change after — is the heart of durable SKAN measurement.

The honest, on-topic place an agency adds value here is the heavy lifting: server-side postback handling, conversion-value schema design and MMP integration. MB Adv Agency builds that layer end to end as server-side tracking, reconciles cross-channel attribution discrepancies through a Google Ads PPC audit, and feeds the result back into the reported metrics defined in Apple Ads performance metrics and reporting.

What Actually Changed in 2026 — and What Did Not

Apple Ads attribution is the platform’s highest-drift topic, so it is worth stating plainly what moved recently and what stayed put. Three dates carry most of the weight: iAd shut down February 7, 2023; Apple Ads added view-through attribution on March 27, 2025; and Apple Ads registered with SKAdNetwork v1–3 on April 10, 2025 — its first-ever SKAN participation.

On top of those, AdAttributionKit (WWDC June 2024) continues to mature — WWDC 2025 (iOS 18.4) added configurable view-through windows and cooldowns — while staying a roadmap item for Apple Ads rather than a live integration. And on the reporting side, Apple’s Custom Reports were absorbed into the new Insights analytics engine in March 2026; the attribution data itself still comes from AdServices and SKAdNetwork, so Insights changes where you read the numbers, not how Apple computes them.

What did not change is the load-bearing model: two parallel paths, deterministic first-party plus anonymized cross-network, with ATT gating neither. MB Adv Agency treats that stability as the point — the framework names and dates drift, but the architecture is constant, so a setup built on the two-paths model survives the next rename. The cost levers that pair with this measurement work live in how much Apple Ads cost.

How to Set Up Apple Ads Attribution

Durable Apple Ads attribution is a sequence, not a single toggle. The order matters: capture the first-party signal, secure the cross-network path, make a deliberate ATT decision, then unify everything and add growth context.

1. Integrate the AdServices framework (iOS 14.3+). Add AdServices so the app can capture the attribution token through AAAttribution.attributionToken(). This is the on-device source of all Apple-attributed, keyword-level data.
2. POST the token to Apple’s Attribution API within its 24-hour TTL. Forward the token server-side to Apple’s Attribution API inside the 24-hour window to retrieve campaign, ad-group, keyword and creative data. Miss the window and that install has no Apple-attributed record.
3. Confirm SKAdNetwork registration for anonymized cross-network postbacks. Verify SKAdNetwork is in place so Apple Ads can be compared against other networks, and plan for AdAttributionKit as the successor.
4. Decide on the ATT prompt deliberately. Apple Ads attribution works without it; ATT only adds the Detailed-payload engagement timestamp. Drive the prompt decision by your other networks’ needs and your conversion-rate tolerance, not by Apple Ads measurement.
5. Connect a mobile measurement partner. Wire up AppsFlyer, Adjust, Singular, Branch or Kochava to unify AdServices + SKAN in one place and add the post-install LTV and ROAS that Apple reports for neither path natively.
6. Link App Store Connect App Analytics for acquisition context. Connect App Analytics to read first-time downloads, redownloads and total downloads, tying paid activity to overall app growth.

MB Adv Agency runs this sequence on every new Apple Ads account before optimizing a single bid, because attribution wired wrong cannot be optimized toward the right outcome. We build the capture and postback layer through server-side tracking; if your Apple-attributed and SKAN numbers look broken rather than merely different, get in touch for an Apple Ads attribution and tracking review.

Apple Ads Attribution & Privacy FAQ

Methodology

This pillar consolidates three absorbed Apple Ads glossary pages — how attribution works in Apple Ads, privacy changes and their impact on Apple Ads campaigns, and integrating Apple Ads with Apple Analytics. Every framework name, version and date is sourced to Apple’s own Help and Developer documentation, verified June 30, 2026: App ad attribution overview (0094), Measuring ad performance (0028), AdServices, SKAdNetwork, AdAttributionKit, and App Tracking Transparency. Named mobile measurement partners — AppsFlyer, Adjust, Singular, Kochava, Branch — are cited only for integration and interpretation color, always attributed. Search-volume data is Ahrefs, June 2026. No MB Adv Agency client metrics appear here; the agency point of view is qualitative, because mbadv has no Apple-Ads attribution benchmark dataset. The load-bearing correction throughout: Apple Ads registered with SKAdNetwork versions 1–3 on April 10, 2025, not SKAN 4. Reviewed by MB Adv Agency, June 2026.