Sports Brand × AppsFlyer · MARS Attribution · Feb 2022

Paid Media
Attribution

A comprehensive assessment of paid media attribution challenges, iOS 14.5+ ATT impact, and a strategic roadmap for privacy-durable measurement across the client app ecosystem.

83%

iOS ATT Opt-Out

30%

EU Users Impacted

50%

US Users Impacted

96%

SKAN via Facebook

01 · Context

The Attribution Landscape

Industry-wide shifts are fundamentally breaking traditional paid media attribution. The client faces compounding challenges across three dimensions simultaneously.

83%

iOS ATT Opt-Out Rate

83.15% of total iOS users opted out of tracking. SKAN not yet configured to pass back revenue.

37–49%

Visitors Without 3P Cookies

37% EU and 49% US client visitors from browsers where 3PC are blocked by default.

96%

SKAN Installs via Facebook

Facebook captures 96% of all SKAdNetwork-attributed installs — critical platform concentration.

Display as Last Click

Only 3% of conversion paths have Display as last interaction. Display is a top-of-funnel assist.

Industry Drivers

Three Forces Breaking Attribution

🍪

3rd Party Cookie Depletion

Safari and Firefox block 3PC by default. Chrome allows blocking. GMP is an SRN — receives postbacks from AppsFlyer but reports its own attribution, creating systematic discrepancy when 3PC is absent. 35–50% of EU/US web traffic not tracked by 3PC. 10% of Chrome traffic is Android webview in EU.

Critical · Now

📱

Mobile Advertising ID Depletion

Apple ATT requires apps to request IDFA permission. Only 16.85% opt in — with 96.23% iOS 14.5+ adoption the impact is near-universal. Google testing AAID opt-out and Topics API. SKAN required for the 83.15% who opt out of IDFA but attribution of revenue is still WIP.

Critical · Now

🔗

Persistent Customer Identifier

Persistent ID across fragmented user journeys is key to cross-channel attribution. The simultaneous collapse of IDFA and 3PC removes the two primary stitching mechanisms. iCloud Private Relay further anonymises Safari users — 19% of Safari users (35%/49% EU/US) can enable it, distorting GMP's probabilistic attribution.

Strategic · 2022–2024

02 · Attribution Flows

Scenario Mapping: Google & Apple

How attribution is assigned across different user journey paths — and where paid media credit is lost to organic or goes unmeasured.

Google Ecosystem

Scenario A
ORGANIC

Ad Click

→

Desktop
Website

→

App Store

→

Install /
Purchase

Organic ✗

Scenario B
PAID ✓

Ad Click

→

Mobile
Website

→

App Store

→

Install /
Purchase

Paid Media ✓

Scenario C
PAID ✓

Ad Click

→

Mobile
Website

→

App Install
Deep Link

Paid Media ✓

Scenario D
IN-APP AD

In-App Ad

→

Install /
Purchase

Paid Media ✓

AAID on Android

Apple Ecosystem — ATT Complications

Scenario A
ORGANIC

Ad Click

→

Desktop
Website

→

App Store

→

App Install

Organic ✗

Scenario B
PAID ✓

Ad Click

→

Mobile
Website iOS

→

App Store

→

App Install

Paid ✓

SKAN if ATT opt-out

Scenario C
AT RISK ⚠

Ad Click
iOS 14.5+ ATT

→

Mobile Web
3PC Blocked

→

Install
No IDFA

Unattributed ⚠

Revenue not passed back

Key finding: AppsFlyer underreports paid media attribution between web and app. Rising opt-out rates require SKAN configuration — currently WIP. 19% of Safari users can enable iCloud Private Relay, anonymising device and IP signals used by GMP for probabilistic attribution.

03 · ATT Impact Data

iOS 14.5+ Decimated Paid Attribution

Measured impact across EU and US markets. Paid channels: Apple Search, Google Display & Search, Facebook, Snapchat. Data: AppsFlyer Apr 2021 – Jan 2022.

30%

EU Client App Users Impacted by ATT

MAID share declined 30% across Europe in 2021. As iOS 14.5+ spread through the EU user base, paid install and revenue attribution declined in lockstep.

50%

US Client App Users Impacted by ATT

The US market saw faster and larger MAID depletion — 50% of US client app users impacted. US iOS 14.5+ adoption was faster, front-running the EU curve.

EU · Paid Installs by Apple Declined

As iOS 14.5+ adoption spread across the EU, paid installs attributed to Apple channels declined directly in correlation. Each percentage point of iOS 14.5+ adoption removed IDFA from measurement, reducing AppsFlyer's ability to deterministically attribute installs to paid campaigns across Apple Search, Google, Facebook, and Snapchat.

Apr 2021 – Jan 2022 · EU

EU · Paid Revenue by Apple Declined

Revenue attribution followed the same decline curve as installs. Without IDFA, AppsFlyer cannot deterministically match ad click to in-app purchase. SKAN-based modelling carries significant accuracy limitations and is not yet configured to pass back revenue data — creating systematic iOS ROAS underreporting.

Apr 2021 – Jan 2022 · EU

US · Paid Installs Decline (Steeper)

The US market saw a steeper decline due to earlier iOS 14.5+ adoption. 50% of US client app users impacted vs 30% in EU — but EU trajectory suggests convergence. The gap between Android attribution (reliable via AAID) and iOS attribution (broken without SKAN revenue config) is widening across both markets.

Apr 2021 – Jan 2022 · US

US · Paid Revenue by Apple Declined

Same revenue attribution collapse in US. Combined IDFA loss and absent SKAN revenue passback means significant iOS-driven in-app purchase revenue is invisible to paid measurement. DV 360 note: 83.15% iOS opt-out, SKAN not passing revenue, 19% Private Relay risk, 35–50% no 3PC, 3% Display last click.

Apr 2021 – Jan 2022 · US

96.23%

iOS 14.5+ Adoption Rate

Near-universal adoption ensures the ATT opt-out problem is not temporary — it is the new baseline for iOS measurement.

16.85%

ATT Opt-In Rate (All Markets)

Only 16.85% of iOS users opted into IDFA tracking. Data: AppsFlyer all markets, Jan 2022.

19%

Safari Users Enabling Private Relay

19% of total Safari users (35%/49% EU/US client visitors) can enable iCloud Private Relay — anonymising IP from GMP.

04 · Problem Statement

Client Attribution Challenges

The combination of industry signal loss and client-specific configuration gaps creates a compounding measurement problem across the entire paid media operation.

Challenge · Critical

Web-to-App Attribution Underreporting

AppsFlyer underreports paid media attribution between web and app journeys. When a user clicks a paid ad on desktop, visits the website, then installs the app separately, the install is attributed to Organic. Server-side solutions are required for robust web-to-app attribution. Smart Banners require mobile SDK on the website; OneLink requires developer deep-link implementation in the client app.

OneLink Smart ScriptSmart BannersServer-Side S2S

Challenge · Critical

SKAN Not Configured for Revenue

Rising ATT opt-out rates (83.15%) require SKAN for iOS paid media attribution, but SKAN is Work In Progress. Not yet configured to pass back revenue data — iOS in-app purchase revenue attributed to paid media is invisible. Only 16.85% of iOS users provide IDFA. The remaining 83% require SKAN for any attribution at all, but currently only install count is captured.

SKAN WIPConversion Studio6-bit Revenue Schema

Challenge · High

Tealium Event Streams Not in EMEA Yet

Multiple solutions depend on Tealium Event Streams being active in EMEA. PushAPI raw data streaming, offline conversion import to CM 360, and persistent cross-channel view all require Tealium as the underlying data infrastructure. mParticle is the current alternative under evaluation alongside GA4. Workflows between mParticle / Tealium / GA4 / AppsFlyer need to be explored.

Tealium EMEA RolloutmParticle AlternativeGA4 WIP

Challenge · High

No Persistent Customer ID

ECID is a 3PC-based identifier with a limited lifespan. Without a persistent cross-channel customer ID, cross-channel attribution is broken. Click IDs (GCLID) are available when present. GDPR obfuscation reduces match rates for User IDs. The simultaneous collapse of IDFA and 3PC removes both primary stitching mechanisms — server-side collection and Tealium connectors are the path forward.

CUID IntegrationGCLID PassthroughPeople-Based Attribution

Challenge · Medium

GMP Attribution Discrepancy

GMP is a Self-Reporting Network — receives postbacks from AppsFlyer but reports its own attribution. Without 3PC, GMP uses probabilistic modelling which isn't deterministic and creates systematic discrepancy between AppsFlyer and CM 360. iCloud Private Relay further anonymises Safari users' IP and device signals which GMP relies on for probabilistic matching. 37%/49% of EU/US client visitors are from 3PC-blocking browsers.

LinkID APICM 360 Offline ConversionHybrid Setup

Challenge · Medium

Android vs iOS Budget Blindspot

With IDFA largely unavailable and SKAN not yet passing revenue, informed Android vs iOS budget split decisions are very difficult. AAID remains available for now, making Android attribution significantly more reliable. A data science multiplier approach is needed to estimate iOS-driven revenue until SKAN is properly configured and validated against historical baselines. Focus on Acquisition vs Revenue needs resolution.

AAID Available AndroidIDFA 16.85% iOSData Science Multiplier

05 · Potential Solutions

Privacy-Durable Measurement Solutions

A portfolio of complementary solutions to rebuild attribution in a privacy-first world — from immediate tactical fixes to strategic platform implementations.

High Return

Priority 1 · Immediate

AppsFlyer PushAPI Raw Data

Stream raw AppsFlyer attribution data to Google Cloud Storage via PushAPI. CM 360 Data Transfers automated. Raw data report needs GCLID for Search revenue. Tealium Event Streams will provide this free when live in EMEA. Currently a premium solution requiring GCS bucket.

PushAPIGCSCM 360 DT

High Return

Priority 1 · Immediate

People-Based Attribution (PBA)

AppsFlyer PBA shows the full cross-device, cross-channel path to app conversion. Essential for understanding multi-touch reality of client app journeys. Requires CUID or persistent ID — match rates depend on login rate and consent levels across the user base.

PBA ActiveCUID MatchCross-Device

Medium Return

Priority 2 · Q2 2022

SKAN Conversion Studio

AppsFlyer's solution to SKAN's 6-bit limit — custom revenue, event and engagement configuration (2^6 = 64 combinations). Configuration is globally complex and subjective. 2–3 day delay. Does not support S2S. Decode mapping file must be uploaded manually.

64 ValuesRevenue SchemaApp Events

Medium Return

Priority 2 · Q2 2022

Tealium Persistent View

Tealium ad-tools connector to push offline conversions when Click ID or persistent customer ID is available. Creates persistent view across marketing channels. Requires Tealium Event Streams setup. Click ID/CUID/MAID required for higher match rates. GDPR obfuscation reduces User ID match rates.

Offline ConversionClick ID MatchTealium Connector

Beta

Priority 3 · Q3 2022

PredictSK — Predictive Analytics

Privacy-centric ML API identifying correlations between early user engagements and predicted LTV. Can work with ATT opt-out users. Beta — requires CSM liaison. Accuracy evaluation needed. AAID available for Android but only 17% IDFA for iOS. SKAD config must be done properly to receive signals.

Predictive LTVATT-CompliantBeta

Evaluation

Priority 3 · Q3–Q4 2022

LinkID API + DV360 Hybrid

LinkID API for DV 360 expected to provide better measurement. Hybrid advertiser setup — mobile web only, stop targeting iOS or use multiplier. App revenue will not show in SA 360; app campaigns should move to Google Ads. AppsFlyer is last-click; Google uses probabilistic modelling as fallback.

LinkID APIHybrid DV360iOS Multiplier

Adapted Techniques

Adapt App Marketing by Device

🤖

Android · AAID Available

App monetisation with custom Android deep links. AAID still available — deterministic attribution possible. Prioritise Android budget allocation until iOS SKAN is configured and validated against historical baselines.

🍎

iOS · IDFA Limited

App monetisation based on contextual signals and iOS deep links as IDFA limited to 16.85% opt-in. Measurement should use data science multiplier for iOS activity. Focus on retention/re-engagement. Contextual targeting for acquisition.

🔗

OneLink + QR Codes

OneLink custom links and QR codes to drive installs in app campaigns. Mobile web-to-app and app page deep linking. OneLink Smart Script for web attribution. Requires developer input for deep link behaviour implementation in the client app.

06 · SKAdNetwork

SKAN Attribution Deep Dive

How Apple's SKAdNetwork works within the client app ecosystem — from ad engagement through to postback delivery and ad network optimisation.

7-Step Attribution Flow

User Engages with Ad

iOS user sees and clicks paid ad. ATT opt-out status already determined at device level.

App Install & Launch

User installs client app. AppsFlyer SDK embedded fires on first launch. mParticle integration active.

SDK registers iOS SKAN. AppsFlyer records first app open. SKAD config determines conversion schema.

Requires SKAD Config ⚠

In-App Actions + Conversion Value

User takes in-app actions. Conversion value updated per schema (6 bits = 64 combos). Revenue, events, engagement.

iOS delays ~1–2 days

iOS Postback to Ad Network

iOS sends postback to ad network with conversion value. Copy sent to mParticle for decoding.

Results delayed 2–3 days

Ad Network → AppsFlyer

Ad network sends postback to AppsFlyer. Networks optimise using decoded postbacks. SSOT comparison (Beta).

PushAPI → Advertiser

PushAPI forwards postback copies to client for raw data analysis and data science modelling.

6-Bit Conversion Schema

Bits 1–3 · Revenue

First three bits allocated to revenue tiers. 8 revenue buckets (2³) to capture approximate purchase value ranges. Balance granularity against statistical noise of aggregated reporting.

█ █ █ ○ ○ ○

Bit 4 · App Event

One bit for a key app event flag — e.g. "added to cart" or "account created". Binary signal used for optimisation signal to ad networks. Configuring the right event is critical.

○ ○ ○ █ ○ ○

Bits 5–6 · Engagement

Last two bits for engagement level — 4 engagement tiers (2²). Captures retention signal, sessions, or product views. Provides optimisation signal beyond pure revenue for top-of-funnel campaigns.

○ ○ ○ ○ █ █

96%

Facebook SKAN Share

Facebook captures 96% of all SKAN-attributed installs. SKAN data Apr 2021–Jan 2022, all markets.

Max Conversion Combos

SKAN 6-bit limit = 64 unique conversion values. Global configuration is subjective — no universal schema.

Day Max Reporting Delay

iOS delays 1–2 days, SKAN results add 2–3 more. Up to 5 days total delay in performance data.

16.85%

ATT Opt-In Rate

Only 16.85% opt into IDFA. 96.23% on iOS 14.5+. SKAN mandatory for the 83.15% who opt out.

07 · Roadmap & Takeaways

2022 Implementation Roadmap

Subject to change — prioritised by impact and dependency. Three strategic pillars: Be In Control · Be Trusted · Be Remembered.

Q1 2022

Be In Control

SKAN Conversion Schema — Store Kit Ad Network conversion value configuration. Revenue, events, engagement 6-bit schema.

PushAPI Raw Data Streaming — Stream AppsFlyer raw data to Google Cloud Storage for CM 360 automation.

Android vs iOS Budget Split — Data science multiplier for iOS measurement gap while SKAN is configured.

Q2 2022

Be Trusted

LinkID API (DV 360) — Better app campaign measurement. Hybrid advertiser setup for mobile web targeting.

PredictSK Beta — CSM liaison for trial access. ML-based LTV prediction for ATT opt-out users.

Tealium / mParticle / GA4 Workflows — Map data flow between platforms for persistent cross-channel view.

Q3 2022

Be Remembered

Incremental Lift Testing — Measure true incrementality of paid media. Holdout testing methodology.

Engaged Campaigns — Push notifications and remarketing where possible. LTV and retention optimisation.

PredictSK Evaluation — Assess accuracy of predictive scores across Android (AAID) and iOS (SKAN signals).

Q4 2022

Scale & Optimise

Privacy-Durable Measurement — Full privacy-safe attribution infrastructure across all channels.

Tealium EMEA Live — Unlocks free PushAPI, offline conversion import, and persistent cross-channel view.

SSOT Beta — AppsFlyer Single Source of Truth — compare SKAN data from Apple vs AppsFlyer.

Key Takeaways · By Stakeholder

Marketing Action Plan

Client

›

SKAN Configuration — Configure conversion studio values globally. Revenue + app events + engagement schema. Enable af_flag for SSOT.

›

PushAPI Raw Data — Stream raw attribution data to GCS. Ensure GCLID present in raw data report for Search revenue attribution.

›

Platform Workflows — Define mParticle / Tealium / GA4 / AppsFlyer workflows. Clarify SKAN postback routing from mParticle to avoid conversion value decoding issues.

Operations

›

Targeting — MAID-based on Android. Contextual targeting with deep links on iOS where IDFA is unavailable (83.15% of users).

›

Optimisation — Focus on Retention and App Events as optimisation signals. Revenue-based optimisation limited until SKAN passes revenue data.

›

Measurement — Activity Revenue or data science multiplier for iOS. Do not rely on last-click ROAS for iOS without confirmed SKAN revenue config.

Media

›

Android vs iOS Budget Split — Over-index Android until SKAN is fixed and iOS ROAS can be validated against the data science multiplier baseline.

›

Acquisition vs Revenue — If acquisition focus: SKAN install data available. If revenue optimisation: SKAN must first be configured to pass revenue signals before optimising toward it.

›

Privacy-Durable Measurement — Build toward incremental lift, PredictSK, and contextual models as the long-term measurement framework beyond IDFA and 3PC.

Paid MediaAttribution

The Attribution Landscape

Three Forces Breaking Attribution

Scenario Mapping: Google & Apple

iOS 14.5+ Decimated Paid Attribution

Client Attribution Challenges

Privacy-Durable Measurement Solutions

Adapt App Marketing by Device

SKAN Attribution Deep Dive

7-Step Attribution Flow

6-Bit Conversion Schema

2022 Implementation Roadmap

Marketing Action Plan

Paid Media
Attribution