Ongoing cookie deprecation and signal loss have eliminated access to the IDs that have underpinned targeting and measurement in marketing. Safari and Firefox no longer support third-party cookies for ad targeting, and Apple’s Intelligent Tracking Prevention makes it hard to use mobile IDs.
That makes only about 30% of the total online audience addressable and measurable today. When Chrome stops supporting cookies, marketers will be in a lot of trouble.
In response to these changes, brands and agencies are actively exploring data collaboration technologies. The hope is that some combination of clean rooms, first-party data and privacy enhancing techniques can help advertising partners collaborate on data files and identify matches, deliver targeted advertising, measure campaign effectiveness and provide greater data transparency.
Adoption of collaboration tools and moving to a first-party world marks a significant shift for an industry long fixated on tracking using third-party solutions. Yet it’s becoming clear that many advertisers view these tools as a way to maintain the status quo in the face of cookie deprecation and tighter privacy regulations. They don’t necessarily want new solutions or different ways of operating; they’re seeking replacements.
In addition to this misguided desire to find a third-party cookie replacement with as little disruption as possible, confusion reigns about these data collaboration alternatives, their capabilities and the differences between them. Here are some key differentiators to keep in mind:
Sorting through the chaos
“Data collaboration” refers to the broader category that includes data clean rooms, enclaves and encryption technology. All of these tools can help marketers privately and securely analyze, activate and measure marketing data.
Some marketers may need to use all of these tools, but others may only need one or two. To better understand the needs relative to your use case, it helps to understand some of the categories of privacy-enhancing technologies in use today.
For example, it may surprise some marketers that most clean rooms do not provide security guarantees and that not all encryption is equal.
In fact, the very concept of encryption has gotten muddy. Many marketers adopting new tools believe that hashing email addresses qualifies as encryption, but it’s not. Hashed emails are obfuscated, not encrypted.
Techniques such as the salting or hashing of data and tokenization are legacy techniques that have been used by marketing and technology companies for many years. Crucially, the challenge with these techniques is that the data is transferred to the party that the brand wants to collaborate with and may then be decrypted within that party’s data environment.
Differential privacy is another form of privacy-enhancing technology. This tech introduces an error rate or signal noise to the data. When obfuscated, it’s very difficult to turn the data back into the original set.
A challenge for marketers with differential privacy is that it often results in a lack of precision with the analysis, which can matter when measuring and attributing ad campaigns. Imprecise data matching matters even more outside of marketing, in verticals such as financial services or health care.
Another technology is federated learning (FL). Rather than bring all data to a central processing point, FL keeps data on its native device and passes a model of that data to the processing point. FL requires a great deal of trust in the company building the model and pushing the data back out. It’s also only viable for large organizations with truly massive amounts of data, such as Google.
A more hardware-based approach is trusted execution environments, also known as enclaves or confidential computing. In short, enclaves deploy a memory-in-hardware-based approach. Think of an enclave as a castle that stores data. Of course, even castles can be sieged, and data can be lost in enclave breaches. Enclaves also require brands to trust the company operating the hardware. This means brands can get locked into one type of enclave, such as a cloud provider, making it difficult to work across other clouds.
Finally, we come to solutions such as fully homomorphic encryption and secure multiparty computation. Both offer full encryption, whereas a lot of the other techniques only encrypt data during the hosting/storage and the processing stages. That means these technologies can offer security guarantees. The historic challenge with full encryption solutions has been the slow speed of computation and therefore the cost. It could take hours of computing time and power to collaborate on fully encrypted data sets, and that often means astronomical computational costs.
A better future through understanding
This information isn’t shared to further confuse brands or bog them down in more terminology, but to explain the options, use cases and offer some means of stratification. No solution is a perfect continuation of cookies. Nor is any one solution perfect for every single marketing organization.
Armed with the deeper knowledge, brands can see that data collaboration needs to be at the heart of their strategy, rather than just a Band-Aid or continuation of the past.
“Data-Driven Thinking” is written by members of the media community and contains fresh ideas on the digital revolution in media.
Follow Pyte and AdExchanger on LinkedIn.
