adidas × MARS · Privacy-First Marketing · March 2022

Data Clean Rooms

Privacy-First Marketing is the new standard — involving ethical and secure data collection, processing and measurement. Can privacy-first and data-driven marketing truly co-exist?

Walled Garden Ad Revenue Share

76% of digital ad revenue driven by Google, Meta & Amazon — creating dangerous dependencies

GDPR Breach Penalty

Up to 4% of global annual turnover or €20M — privacy compliance is now existential

2022

Year of Privacy-First Reckoning

3PC deprecation, IDFA collapse, AAID restrictions — every identity signal under pressure

Regulatory Environment

The Regulatory Pressure

Data protection laws are setting the floor for privacy-first marketing — and the penalties for non-compliance are severe enough to redefine business strategy.

United Kingdom

UK Data Protection Act

Permission required from online users before collecting data for marketing
Clear explanation of data usage in owned and operated media is mandatory
Data storage, processing and measurement must happen in privacy-first environment

⚠ Breach Penalty: 4% turnover or £17.5M

European Union

GDPR

Permission required from online users before collecting data for marketing
Clear explanation of data usage in owned and operated media is mandatory
Data storage, processing and measurement must happen within EU in privacy-first environment

⚠ Breach Penalty: 4% turnover or €20M

Identity Deprecation Timeline

A Decade of Signal Loss

The systematic dismantling of web and mobile identifiers — from ITP 1.0 in 2017 to ATT in 2021 — has fundamentally broken traditional attribution.

2016

iOS 10+ LAT — Limit Ad Tracking

iOS 10+ users allowed to opt out of tracking via Limit Ad Tracking (LAT) setting. First major signal of Apple's privacy trajectory.

IDFAiOS

2017

Safari ITP 1.0 — 3PC life limited to 1 day

Intelligent Tracking Prevention caps third-party cookie lifespan. Safari, representing 35–50% of adidas web traffic, effectively removes 3PC.

Safari3PCITP

2018–19

ITP 1.1 → 2.2 · Storage API · 1PC limited to 1 day

Progressive ITP updates eliminate workarounds. ITP 2.1 blocks 3PC entirely on Safari. ITP 2.2 limits first-party cookies to 1 day in some contexts. Mozilla Firefox ETP 1.0 allows opt-out of 3P tracking.

SafariFirefoxITP 2.x

2020

Firefox ETP 2.0 · Chrome Samesite Update · Chrome 3PC Delay

Firefox ETP 2.0 blocks 3P tracking by default. Chrome introduces SameSite cookie flexibility but also announces 3PC deprecation — later postponed to 2023.

FirefoxChromeSameSite

2021

iOS 14.5 ATT · iOS 15 Hide My Email · iCloud Private Relay

App Tracking Transparency requires explicit opt-in for IDFA. Only 16.85% of adidas users opt in. iOS 15 adds IP cloaking and Hide My Email. iCloud Private Relay enables 19% of Safari users to anonymise IP from ad servers. Android AAID opt-out announced.

ATTIDFAPrivate RelayAAID

Walled Garden Ad Revenue

Google, Meta & Amazon dominate digital ad spend — now is the time to test privacy-centric alternatives

0.85%

ATT Opt-In Rate

Only 16.85% of iOS users opted into IDFA — 96.23% of users now on iOS 14.5+

0–49%

Visitors Without 3PC

37% EU / 49% US adidas visitors from browsers where 3PC are blocked by default

Max GDPR Penalty

4% of global annual turnover or €20M — data breaches are now a board-level existential risk

Technology Explained

What is a Data Clean Room?

A privacy-first data sharing infrastructure enabling encrypted, aggregated datasets from different vendors to be compared with client CRM data — without any data leaving its origin.

Definition

Privacy-First Data Collaboration

DCR provides a privacy-centric infrastructure which allows encrypted aggregated data sets from different vendors to be shared with client CRM data or between themselves for mutual benefits.

This does not provide omni-channel attribution. It is not a substitute for persistent identifiers — rather a complement that works within privacy constraints.

Differential Privacy Aggregated Data Nascent Technology Not Omni-Channel Not an ID Replacement

Key Use Cases

Primary Applications

The key use case is Matched 1PD activation, suppression and measurement. Keys matched include: City, Email, First name, Gender, Last name, Phone, State, ZIP code and other hashed identifiers.

Enables brands to activate CRM segments against publisher audiences, measure incremental sales, suppress existing customers, and enrich profiles — all without raw data exposure.

1PD Activation Audience Suppression Incremental Measurement Profile Enrichment Media Planning

Types

Media DCR vs Neutral DCR

Two Distinct Architectures

Neutral DCR

Media Agnostic

Example: Habu · Infosum

Provided by neutral partners — not owned by media co.
Matches client 1PD with different data & media vendors
Matched data sets pushed to different media tools for activation
Media agnostic — works across platforms
Suppression doable across ecosystem
Not an omni-channel attribution solution

Media DCR

Ecosystem Locked

Example: Google BigQuery Omni · AMC

Matches client 1PD directly with media vendor data sets
Deep integration with platform-native insights
Suppression doable only within vendor ecosystem
Owned by media partner — not neutral
Matched data can ONLY be used within vendor's environment
Not a media agnostic solution

🏛️

Infosum

Bunker Technology · Immovable Data

Provides infrastructure with Bunker technology to connect, analyze and activate — ensuring data never moves from its origin. Leading advocate of immovable data sets with Bloom filters (hashing) in their data matching approach to introduce deliberate noise.

Commercial trust through decentralized architecture
Increased compliance — data never leaves its bunker
Data normalization across client, publisher, and data bunkers
Decentralized — no central data repository
Partnerships across key UK/EU media owners
Friendly user interface — easy roll-out at low cost

⚠️

Limitations

Scorecard Assessment

While Infosum excels at privacy-first data matching, its analytical capabilities are more limited than Habu — particularly for advanced ML and predictive analytics use cases.

Limited Advanced Analytics — little functionality to analyze data
No Predictive Analytics capability
No connectors with Walled Gardens (Meta, Google, Amazon)
Infosum is not fully federated
Limited customisation possible
ID graph not available for EU markets

SWOT

Infosum Analysis

↑Strengths

What Infosum Does Well

Friendly user interface — accessible to non-technical users
Valuable for clients with sensitive data requirements
Roll-out is easy and low-cost to implement
Strong partner integrations across UK/EU media owners

↓Weaknesses

Capability Gaps

Limited Advanced Analytics and reporting depth
No Predictive Analytics or ML modelling
No direct connectors with Walled Gardens
Not fully federated — limited cross-system interoperability

◎Opportunities

Near-Term Pipeline

Tealium Event Streams integration — WIP
Ozone ad network partnership — WIP
SkyTV will be added as partner soon
The Trade Desk (TTD) available from Q2 2022

⚠Threats

Market Risks

Widespread Privacy by Design with integrated secure match capabilities could make the offering redundant
Not interoperable with other DCRs or cloud tech stacks

Platform Deep Dive · DCR Suite

Habu — Interoperable Suite

DCR software providing distributed data connection and advanced analytics with AI/ML. Built on Snowflake — cloud neutral, schema-on-join architecture that preserves 100% of source data value.

Integration Layer

Foundation

Cloud Sources Media Ecosystems Neutral Cloud Stance Zero Data Movement APIs into Walled Gardens

Application Layer

Control Plane

Setup & Provisioning Join & ID Resolution Privacy & Governance Controls Reverse ETL / Segmentation Activation & Audience Distribution Code Execution

Intelligence Layer

Analytics & ML

Automated Queries & Reporting Advanced Modelling / ML Audience Insights Measurement & Attribution Distributed Machine Learning

First Generation DCR Problem

Data Devaluation on Upload

Traditional DCRs require data to be uploaded and mapped to a global schema — normalizing and devaluing data from source. Hard to map back to log files or accurately run machine learning. Data is devalued from source taxonomy before it can even be used.

Habu's Schema-on-Join Advantage

100% Data Value Preserved

No predetermined taxonomy or schema means data stays in its native form. 100% of data is available for collaboration use cases. Processes all formats including ad logs, ML models. Always fresh data sourced directly from cloud sources — never stale copies.

SWOT

Habu Analysis

↑Strengths

Habu Advantages

Templates for Advanced Analytics — ready-to-use models
Highly secure as built on Snowflake infrastructure
Fully Interoperable — partners need not be on Habu
Curated Business Intelligence and reporting

↓Weaknesses

Commercial Considerations

SaaS license model requires upfront financial commitments
Starter Packs available for cloud config like ADH or AMS
Higher complexity than Infosum for simple use cases

◎Opportunities

Strategic Fit

Customizable for existing workflows and ad-tech stacks
Global choice — seamlessly works across regions using different DCRs
WPP is on Snowflake — Habu products integrate natively

⚠Threats

Competitive Risks

If clients have in-house Data Scientists, they may adopt Blockgraph or Snowflake natively
Internal measurement teams may prefer direct cloud DCR (ADH/AMS) over SaaS model

Analytics Capability

Tailor-Made Analytics Suite

A full-funnel advanced analytics toolkit — from brand awareness measurement at the top, to conversion and incrementality at the bottom. Phase 1 signals enable an ever-expanding capability set.

Upper Funnel · Awareness

Reach & Frequency Analysis
Share of Voice
Media Mix Modelling
Customer Journey Mapping

Mid Funnel · Engagement

Campaign Performance
Propensity Modelling
Causal Impact Analysis
Audience Segmentation

Lower Funnel · Conversion

Attribution Modelling
Incrementality Analysis
Incremental Sales Measurement
ML-Driven Return Propensity

Incrementality Methodology

Measuring True Sales Uplift

Using Brand Insight Bunker (including purchase data) matched against Media Owner Insight Bunker where control & exposed groups are randomly defined:

Four segments to calculate incrementality:
1. Users exposed to ad who made a purchase
2. Users NOT exposed to ad who made a purchase
3. Total users in control group
4. Total users exposed to campaign

Incremental uplift = Purchase rate (exposed) − Purchase rate (control)

Habu ML Training

Distributed Machine Learning

A global CPG is creating a data flywheel using Habu's ability to securely connect machine learning models to scaled consumer & transaction data:

Brand / Agency benefits: Track cannibalisation, optimise trade spend budgets, run targeted co-marketing

Retailer benefits: CPG partner retention, secure promotions, return propensity scores shared via clean ML environment without raw data exposure

Next Steps

Inform, Discuss & Explore

Bringing together conversations across the business — the purpose of this session is to plan a roadmap to future-proof our data capability.

Recommendation

Now Is the Time to Test

Now is the time to test privacy-centric solutions that work across the fragmented ad ecosystem. The walled gardens capture 76% of digital ad revenue today — Data Clean Rooms provide a path to privacy-compliant measurement and activation that reduces this dependency.

Both Infosum and Habu are valid — Infosum for simpler, cost-effective 1PD matching; Habu for advanced analytics, ML, and full Snowflake interoperability.

Pilot Infosum Evaluate Habu Tealium Integration 1PD Strategy

Discussion Points

Stakeholder Conversations

Key topics to align across business stakeholders:

Stakeholder alignment — bring together IT, Media, Data Science, and Legal teams on DCR strategy

Timelines — align on Q1–Q2 2022 pilot schedule given TTD availability in Q2 and SkyTV partnership imminent

Commercial model — evaluate Infosum (lower cost, simpler) vs Habu (SaaS commitment, higher capability) based on use case roadmap

Privacy-First is the New Normal

The deprecation of identity signals and tightening regulation is not a future threat — it is the present reality. Data Clean Rooms provide a privacy-compliant infrastructure to future-proof measurement and activation.

Data Clean Rooms Privacy-compliant 1PD matching

Infosum Bunker tech · UK/EU partnerships

Habu Snowflake ML · Full interoperability

Contact rahul.upadhay@mediacom.com