How OMG plans to tackle signal loss of cookies with data clean rooms

As a small portion of third-party cookies depreciate from today (4 January), data clean rooms (DCRs) promise to reconnect data that advertisers can use for targeting, personalisation and measurement.

While walled gardens like Google and Meta have dominated DCR activity, Annalect, Omnicom Media Group's data and analytics division, recently expanded its partnership with InfoSum, a data collaboration tech provider, to develop a neutral DCR service for Asia Pacific clients.

Both partners hope their collaboration will bring about more sustainable, effective, and privacy-safe data collaborations, addressing the challenges of fragmented data sources and data silos.

Campaign previously explored how the demand for DCRs has skyrocketed due to privacy concerns and growing first-party data use by advertisers.

We find how this partnership will help OMG's clients from Paul Shepherd, chief investment officer and president of Annalect APAC at Omnicom Media Group; Ravikumar Shankar, chief data solutions officer for Asia Pacific at Annalect and Richard Knott, general manager for Australia and New Zealand at InfoSum.

The three executives also discussed with Campaign how they will address DCRs' limitations.

Why use a DCR?

After building data management platforms (DMPs) for its clients across the region, OMG realised that many DMPs fell short, particularly in creating a unified consumer profile and understanding consumer insights.

Unable to create a unified consumer profile and understand consumer insights has led to a significant gap between the expectations set by DMPs and their actual delivery, which some media networks like OMG have struggled to bridge.

The media agency group's primary goal was to find a privacy-secure method to use data to build insights and activations and understand consumers for measurement purposes.

Initially, Omnicom collected data on a centralised platform from client-owned assets, gaining a good understanding of consumers from the client's viewpoint. However, they lacked insights into the consumers' activities beyond interactions with the client's app or website.

Omnicom sought additional data from other vendors to fill these gaps, but this approach raised data integrity and privacy issues exacerbated by regulations like GDPR. The physical transfer of data between databases was risky and unsustainable.

"We aimed to achieve data granularity, depth, and breadth of data attributes to fully understand the consumer without hosting the data in our environment. We sought a safe way to connect the data, which led us to work with InfoSum," explains Shankar.

Knott adds that whilst there might be challenges regarding the scale of the number of profiles, there is no limit to the potential number of attributes assigned to these profiles.

For instance, there might be only a few hundred thousand profiles in a database, but it is possible to append thousands of attribute columns.

"Once there is a more prosperous and deeper understanding of these profiles, it becomes straightforward to use affinity and lookalike tools. These tools help find people who closely resemble your audience but are not in your database," explains Knott.

"They might watch the same TV shows, purchase similar products, visit the same places for lunch, or follow the same commuting routes. You can effectively scale your targeting efforts by identifying individuals who engage in similar activities but are not part of the initial data set. Essentially, you're targeting people identical to those already in your database."

Addressing interoperability and technical barriers in DCRs

One significant issue with DCRs is that different DCRs cannot be interoperable. There is also a natural technical barrier when using them as they do not have a user interface, requiring these same users to master technical languages like Structured Query Language (SQL).

It is, therefore, essential to have a system that can integrate with various apps and upstream data cloud services like Amazon, Google, and Microsoft. Additionally, the technology should work seamlessly with downstream partners, including customer data platforms, DMPs, and demand-side platforms.

Shepherd acknowledges this, noting that platforms like Google or Meta operate within its unique ID space. They are responsible for protecting their users' data, which they do by creating protective walls around it, hence the term 'clean rooms'.

Through Omni, OMG's marketing and insights platform, the media agency group has developed interoperability through its coding and resources to bridge gaps between these platforms. One of the initial steps in deprecating identity linking without revealing too much is taxonomy.

For example, a user's behaviour on YouTube might involve watching fitness content, while on Facebook, they might engage more with interactive content. OMG is not linking users directly; the agency connects behaviours through a shared taxonomy.

"As an ex-programmatic professional, I've seen a shift from precision marketing to what we now call predictive marketing. Utilising these individual clean rooms as seeds of insight is crucial for enhancing predictive capabilities. When advising clients, it's essential to have capabilities across all platforms and connect them with tools and processes to drive these strategies forward," explains Shepherd.

"In discussions about data management, Google and Facebook are often the primary focus, as they pioneered the original clean room concept. However, the emergence of neutral clean rooms, such as Infosum, presents a new opportunity. These neutral cleanrooms allow for the connection of like-minded advertisers and publishers, but not in the traditional ID space."

Shepherd adds: "Infosum operates differently, not relying on ID spaces, an essential aspect. This trend is gaining traction in the US, where publishers and advertisers increasingly seek a neutral platform that still offers compliance, security, and support."

For example, Infosum's solution is decentralised, meaning data processing occurs within each company's architecture without centralising or commingling data. This approach fundamentally differs from solutions that concentrate data, making interoperability challenging.

Infosum uses direct matches against multiple identifiers for identity management, avoiding reliance on an ID spine or data brokered by Amazon Marketing Cloud (AMC) or other entities. This method ensures data privacy and security, enabling effective data matching and analysis.

Shankar is hopeful that the concept of interoperability will evolve and will no longer be about connecting data but rather about linking insights.

He explains that the agency can use a user's behaviour in one ecosystem as a basis for interoperability in another. This approach will be particularly relevant in different 'clean rooms' like Google, Meta, or AMC.

"Interoperability will focus more on insights rather than data. For example, providing a first-party data set to Google will offer specific insights about my audience," says Shankar.

"If I take the same data set to Meta and AMC, each will provide different insights based on the same set of first-party audience data. This process helps create an 'insight canvas', where insights from various platforms are compiled in diverse ways. This method of joining interoperable systems in the future is crucial because anything related to ID linking is not privacy-safe and is becoming increasingly unfeasible. Therefore, we must develop a new approach focusing on insights rather than direct data connections."

Activating in a DCR

In the context of campaign activation, a key challenge is providing teams with actionable insights for making real-time changes within a campaign. Traditionally, cleanrooms require SQL expertise, limiting their accessibility.

However, Omni has created a SQL library that produces visualisations, offering real-time insights for campaign optimisation. The development of standard queries is particularly relevant in the current landscape where the industry is moving towards 'doing more with less'.

The concept of identity deprecation, like Apple's initiation of App Tracking Transparency, transmits fewer signals in the bitstream. Despite this, robust, compliant, and safe signals are still available to be optimised.

Omni has created an ecosystem that helps teams access and interpret these insights, enabling them to make informed adjustments in real-time. This approach uses available data, enhancing campaigns' efficiency and impact.

"As a trader, I can now easily view my YouTube campaign on the screen. I can check the top-line media spending, my CPM, cost per view, and what's in my funnel. I can analyse the time of day and even question if late-night advertising, like the 7.4% of my budget spent between midnight and 4 am for a Mercedes Benz campaign, is compelling," explains Shepherd.

"Previously, this would have involved cumbersome standard reporting with megabytes of data and pivot tables, but now, I can access all this information with just a click. I can investigate specific details like time of day, advertiser, weekly performance, device costs, reach and frequency, and optimal frequency."

Shepherd explains Shankar has created about 30 different data views, allowing traders to make informed decisions quickly. Previously, traders would have to ask Shankar about the agency's spending on platforms like PubMatic, Rubicon, or DV360, which could take up to 15-20 minutes.

Shepherd adds that Infosum offers a user interface where the agency can orchestrate its data, allowing OMG to build its audiences without writing SQL queries.

"We can now easily understand insights and build audiences. I believe it's the combination of what we've developed in Omni and the native interfaces of platforms like Infosum that enhances our capabilities," says Shepherd.

Knott notes one of the challenges people have encountered with DCR is understanding the taxonomy. It reminds him of that Spider-Man meme where everyone points at each other, saying, "You're a data clean room; I'm a data clean room."

Pointing to how everyone claims to have a DCR, Knott says a DCR is merely a feature, not a complete product. The real value lies in what people build on top of their DCRs.

For example, while Infosum has a DCR, its product is a data collaboration platform where users can drag and drop different datasets, known as 'bunkers', and perform collaborations. They can then generate reports and activations from it, leveraging the DCR.

"The DCR is like the engine, while the data collaboration platform represents the entire car, explains Knott.

"Some say they can't get much out of a DCR. This is often because there isn't a comprehensive product built around it, or they haven't developed anything on top of it. That's where the real challenge lies when people discuss DCRs."

Developing a holistic measurement framework

There is currently a host of DCRs available in the form of Google Ads Data Hub, Facebook Advanced Analytics, AMC, Infosum, Snowflake, LiveRamp, and others.

Most of Annalect's measurement work, particularly in the Google Ads Data Hub, is due to the client using Google Campaign Manager as their ad server, allowing the agency to incorporate cross-media information for measurement purposes. 

For example, the agency can access YouTube and DV360 delivery tables and include all its premium guaranteed delivery with Mediacorp and Carousell. In addition, the agency can bring in information from tracking clicks on Meta.

In the DCRs mentioned above, agencies can build a measurement framework and e-commerce capabilities in AMC.

"You might have one clean room that tells part of the story, but when you want to delve deeper, you need to use the individual components of each platform. If your measurement framework is just about measuring success against some baseline metrics, you can manage that," explains Shepherd.

"A clean room is just the starting point; what you build on top of it matters. When you want to go deeper into the individual platform, you need to utilise the specific features of each particular component."

Shankar points out clients are becoming increasingly aware that the landscape is changing, and they understand the agency cannot connect cross-channel data into one unified system. 

What the agency can do, however, is connect cross-channel data from Google, insights from Google Ads, insights from Meta, and insights from AMC through Annalect's modelling. 

Shankar explains connecting the cross-channel data does not have to be ID-based, as the agency can still model at an aggregate level. For example, the agency uses regression, cluster, and mapping pairwise analysis methods. Similarly, Annalect can stitch insights from multiple sources into one holistic view that understands the overall campaign delivery.

"It is important to note that this approach is not as precise as ID-based methods, but it is certainly better than having no insights. It provides us with directional insights into cross-channel performance," explains Shankar.

Shankar shares an example of a closed-loop measurement the agency conducted for a FMCG brand in Australia.

He explains FMCG brands often have very little first-party data, as few people visit their websites or use their apps. To generate first-party data, Annalect collaborated with retailers to access data about customers' purchasing habits.

This first-party data includes information on the brands people have bought in the last two months, which the agency then ingests into platforms like Meta, Google, and others.

"This approach allows us to understand how many people were exposed to an ad, how many times they were told, the campaigns they interacted with, and the touchpoints involved. We can then build an ROI and understand the touchpoints," explains Shankar.

"Without this data, we could not deliver or measure ROI for the client. However, with this data, we can create a complete closed-loop measurement, assess the campaign's effectiveness, and plan future campaigns."

Shankar adds: "This data helps us understand where people buy and the sizes and SKUs they prefer and provides a holistic view of consumer behaviour. While the volumetric analysis may not be exact, it at least offers a glimpse into this aspect of consumer behaviour.

How significant will DCRs be in a cookie-less world?

Ultimately, DCRs are just one of the many tools in the toolbox for advertisers to address the deprecation of identity or signal loss, as well as for measurement and audience analysis.

It is also essential to note utilising multiple clean rooms can become a significant financial burden, as the cumulative costs can escalate, making it challenging for organisations to benefit fully from their extensive capabilities.

However, the trend Annalect observes is the increasing necessity for marketers to invest in cloud technology, according to Shepherd. He notes the processing power required in advertising has grown exponentially.

Reflecting on his experience over the years, Shepherd recalls discussing publishers and creating media plans using an ad server. Today, with programmatic advertising, he notes the agency is making hundreds of bids in mere milliseconds and processing terabytes of data, which means processing demands robust cloud infrastructure.

"Moreover, the role of AI – or machine learning – is accelerating this process even further. While there's certainly a cost associated with these advancements, they are essential for efficiency and effectiveness in modern advertising," explains Shepherd.

"Investing in cloud-based capabilities, including clean rooms, is no longer just an option but a necessity for those who want to remain competitive and effective in today's fast-paced digital world."