When is a clean room not a clean room?

For a brief moment, marketers hoped that a combination of a clean room and a Customer Data Platform (CDP) would be all they need to achieve connectivity between their data and the wider ecosystem. However, lofty expectations collided with a complicated reality as vendors, big tech platforms, and media owners launched clean rooms of all shapes and sizes, with significant differences in capabilities and use cases.

To provide an overview, there are clean rooms built into data warehouses that allow users to leverage their data where it is stored, without having to port it outside their own ecosystem. Walled gardens operated by the likes of Google and Disney offer clean room environments for audience matching, measurement, and attribution. Then there are standalone data collaboration clean rooms that allow two parties to privately share, analyse, and activate data; and lastly, query-based clean rooms that offer matching across data lakes and warehouses without the original data being moved.

Confusing? Definitely. However, by breaking down each category of clean room into their pros and cons, we can get a sense of their respective capabilities and how marketers can maximise the value of data clean rooms.
 

Demystifying the crowded clean room landscape

1: Data warehouse clean rooms: advanced capabilities for existing customers

These are the clean room capabilities “built in” to data warehouses such as Snowflake and Amazon Web Services (AWS). Marketers whose data resides in these warehouses can use clean room functions to privately collaborate, permit, and analyse data, with the underlying infrastructure making them ideal for tasks that require comparison of two data sets, such as attribution and measurement.

Snowflake and AWS are highly customisable, which can be a pro or a con depending on the user. Those who have the data scientists and engineers on board to take a “builder” approach can unlock highly bespoke and complex functionality, like cryptographic computing. You can bring in your own algorithms, query sets and identity spine. However, they aren’t designed to be accessible to someone without data expertise, or some sort of technical background.

Another consideration for this category of clean room is the requirement for the data to be stored in the underlying warehouse, which means that both collaboration partners will ideally already be customers of the same service.
 

2: Walled garden clean rooms: simple interfaces for single-platform use

Big tech and media players such as Google, Facebook, and Disney have leveraged their wealth of first-party data to launch clean rooms, allowing their customers and partners to perform data matching within their platforms. By identifying overlapping audiences, marketers can narrow down segments for retargeting and suppression on campaigns run within the host’s ecosystem.

These clean rooms are designed with marketers in mind, and as a result, tend to have user-friendly interfaces and relatively simple data porting procedures. However, data must stay confined within the platform owner’s ecosystem; marketers bring their data to Disney but Disney will never allow its own data outside its walls, thus “walled garden.” This is great if you’re running a campaign on the platform, but can't be used across multiple partners and strategies outside of that environment. 
 

3: Data collaboration clean rooms: intelligent platforms for downstream activation

Data collaboration clean rooms are similar to walled garden clean rooms in that they facilitate the matching of data ported into their environment, with the vital distinction that they are not tied to a single organisation’s ecosystem. Partners can collaborate on these platforms on equal footing, and can take the outputs out of the clean room for downstream activation to any supported partner.

Like their walled garden counterparts, data collaboration clean rooms (or “platforms” - many aren’t branded as clean rooms despite possessing similar capabilities) typically have user-friendly interfaces and will perform some of the data orchestration heavy lifting so that they can be used without deep data expertise.

These flexible tools are well-suited for data enrichment, analytics, and activation, but require both parties to port their data into that particular environment. This can cause issues if one partner is signed up to vendor A, while another uses vendor B. Having to pay for different platforms for different partners can quickly eat into the benefits of data collaboration. In some cases, the platform will allow a “guest,” meaning only one party foots the cost of collaboration. 
 

4: Query clean rooms: platform-agnostic, non-portable data matching

These clean rooms are built around the non-portability of data, enabling both parties to make overlapping or match queries while keeping their data within whatever warehouse or private ecosystem it is housed in. Platform-agnostic query clean rooms are perhaps closest to the original promise of clean rooms, allowing marketers to collaborate without the need to relocate their data.

But keeping data in its original location does require more technical sophistication, as the user is responsible for the data transformation required to prepare their sets for matching. Each party must then perform data orchestration across two separate ecosystems, for whom the categorisation and the format of that data could be holistically or intrinsically different, or a lack of shared identity may result in low or poor overlap.

Query clean rooms are ideal for those who do not want to commit to a single ecosystem, or for organisations where non-portability of data is paramount, but are often most successful across partners who share universal identifiers.
 

No clean room is a cure-all for privacy

Now that we’ve covered the major distinctions between the various types of clean room in the market, it’s important to note that none of them are a privacy panacea. They all allow for the secure and private sharing of data - and even consent signals if available from an audit perspective - but that does not mean that the data itself or how it’s used is magically privacy-compliant because it’s been run through a clean room environment.

Let’s say I consented to a shoe store using my data to personalise my shopping experience. If the store then does a collaboration in a clean room with a sneaker brand who then ports that data into its CDP, it would then live in its ecosystem, one-to-one against my profile. Then, if the sneaker brand, with whom I don’t have a relationship, was to send me personalised emails, I might feel that the shoe store shared customer data I was willing to give them with a brand I personally didn’t trust or have interest in. This is especially relevant if that shoe store never stated in their privacy notice that my data would be shared with third-parties, and opted me in to mailing lists I didn’t subscribe to.

When using clean rooms, both parties must still do their due diligence to ensure that their data, their collaborator’s data, and the output of the collaboration tick all regulatory and user agreement boxes and that consent can be tracked and withdrawn - especially if the output is being activated downstream or ported into environments where data lives one-to-one, such as CDPs. 

If an individual did not consent to their data being used for the purposes of the collaboration, or if the end result of the collaboration is that they are personally identifiable, it doesn’t matter how much their data is scrambled at either end, or how much noise is injected into the analysis. Due diligence is still the responsibility of both parties.

From the flexibility of query clean rooms to cosy but confined walled gardens, marketers have plenty of choice in the data collaboration space (perhaps too much, expect consolidation to make its way through this crowded market soon). However, whatever marketers choose, no clean room is a silver bullet. Responsibility for privacy compliance falls on data custodians to ensure data is used ethically, with respect for user content. We must remember that clean rooms are tools like any other - how they are used, or misused, is up to you.