Retail media has seen explosive growth, with spend expected to reach $160 billion by 2027 according to GroupM. However, serious measurement challenges threaten to constrain that growth. While data clean rooms have emerged as a potential solution, they have limitations that prevent robust cross-channel measurement. Brands need more holistic solutions to quantify incremental impact and optimize spend across retail media networks.
The rapid rise of retail media
Retail media networks (RMNs) allow brands to advertise using retailers’ first-party data. This powers targeted ads both on retailers’ owned channels and off-site.
RMNs have skyrocketed for several reasons:
- Massive ecommerce growth during the pandemic, with online shopping reaching 20.3% of all retail according to eMarketer
- The demise of third-party cookies and identifiers like IDFA
- Closed-loop measurement linking ads to sales thanks to first party data
- Retailers monetizing their data and audiences. Retail media revenue hit $30 billion in just 5 years.
This boom has led to a fragmented environment with numerous downsides:
- Proliferation of new RMNs and ad types. Criteo analysis shows top retailers have at least 16 ad types each.
- Walled gardens that prevent data sharing
- Privacy regulations limiting cross-platform tracking
- Inconsistent measurement methodologies. Incremental found 80% of top retail ad types rely on last-touch attribution.
These factors have created significant measurement challenges for brands and agencies.
The limits of data clean Rooms
Data clean rooms emerged as a way to enable collaboration without sharing raw customer data. Advertisers can match and analyze data from multiple parties to create audiences.
Retailers quickly adopted clean rooms to power retail media networks. By anonymizing and segmenting data, they can safely expose it to brands for targeting and measurement.
However, a data clean room may have limitations:
- They reinforce walled gardens as retailers silo data.
- Brands must join each clean room separately.
- Retailers control permissions and the data shared.
- Adding partners increases complexity.
- They focus on narrow predefined use cases.
This closed nature prevents the flexible analysis needed for optimization. Clean rooms also don’t solve the core problem of inconsistent measurement approaches across retailers.
The pitfalls of current Retail Media measurement
Most RMNs rely on last-touch attribution, assigning each sale to the last ad seen. This approach has major drawbacks:
- It ignores the omnichannel customer journey.
- It fails to account for external factors driving sales. Marketing mix models show these drive 60–80% of sales.
- It lacks predictive power for planning budgets.
Last-touch attribution provides no visibility into cross-channel impact or incrementality linked with ad exposure. Brands and agencies lack the holistic insights needed to quantify impact, optimize campaigns, and plan budgets.
Flaws with common measurement approaches
Beyond last-touch attribution, brands use other methodologies that also fall short:
Incrementality testing: Compares exposed and control groups to estimate incremental lift. But small sample sizes limit segment-level insights needed for optimization.
Marketing mix modeling: Statistically models marketing’s impact on sales. But these models require months of data, limiting agility, and they typically exclude retail media.
Clean room analysis: As noted above, closed clean room design hinders flexible analysis for true optimization.
Multi-touch attribution: Uses rules-based or algorithmic models to share credit across touchpoints. But relies on third-party cookies, lacking needed scale and accuracy.
Platform-level optimizations: Platforms like Amazon optimize bids to hit specified targets. But optimizations are limited to the platform and focused on cost efficiency over incremental sales.
These approaches provide fragmentary insights. None tackle the core challenge — limited shared data prevents cross-platform optimization against real incremental sales impact.
The need for holistic measurement
To optimize retail media, brands need a more holistic approach, with aggregated data. Measurement must:
- Combine online and in-store data for omnichannel view
- Incorporate all marketing efforts beyond just retail media networks
- Flexibly analyze diverse datasets without movement or duplication
- Provide a persistent view of impact over time
- Enable optimization against real incremental sales
- Be scalable across all major retailers and channels
This requires both breaking down data silos and moving beyond constrained methodologies. The solution lies in better technology and processes.
A path forward: neutral measurement environments
To address these needs, emerging solutions use neutral, secure environments to combine datasets. New players apply privacy enhancing technologies to enable better analysis without raw data changing hands.
Key capabilities include:
- Connecting exposure data from all channels
- Ingesting transaction data from retailers, CRM systems, etc.
- Providing flexible analytics against sales impact goals
- Supporting optimization and simulation for planning
- Offering measurement across multiple retailers
- Embedding deterministic identity for accuracy
- Maintaining security and privacy compliance
This approach breaks down data silos that constrain measurement while avoiding moving or pooling data. The result is brands and agencies get the insights needed to quantify impact, simulate strategies, and optimize retail media to business outcomes.
The bottom line
Retail media brings huge potential, but brands have struggled to measure its impact. Data clean rooms have helped but fall short of enabling optimization. To continue driving growth, retail media needs improved technology and processes for measurement. Neutral solutions that provide flexible analytics against sales impact goals offer a path to unlocking retail media’s full promise.
Note: this post was previously published in French on Medium: https://medium.com/retail-media-france/retail-media-pourquoi-les-data-clean-rooms-sont-insuffisantes-a50def8c4449
.webp)