Post-cookie does not mean post-identifiability

December 4, 2023

Authored by Kasper Skou, SVP Partnerships

Post-cookie does not mean post-identifiability

There is a prevalent conviction in the industry that the end of the third-party cookie as a means of cross-site user tracking will also mean the end of user identifiability and, as a consequence, audience targeting at any relevant scale.

This is a misconception, and it’s important for marketers to understand why – and how they can move forward confidently in a cookieless world.

The opportunity behind the “loss”

The digital advertising industry has for two decades relied on the third-party cookie as the main identifier of internet users to facilitate cross-site tracking of behavior with a view to building user profiles, targeting users based on these profiles, frequency capping the exposures a user has to a given ad, and measuring the effects of different marketing strategies on business outcomes to determine the attribution that should be assigned to each of these strategies. It now seems that the “stays of deprecation” we saw from Google have finally come to an end in 2024, eliminating the third-party cookie for cross-site tracking in the most widely used browser across most markets: Chrome.

This significant development is a great opportunity for our industry to rid itself of the crutch that is the third-party cookie and find better ways to identify users. The hope is not to find a single solution analogous to the third-party cookie (this, after all, would again concentrate too much power with the proprietor of that solution.) Rather, we’re seeing a small set of post-cookie user identifiers emerge in the market, and the goal will be to enable interoperability among them. These IDs come in various forms, including:

  • Deterministic IDs such as hashed email and its various deployments, for example Unified ID 2.0 (UID2, or EUID, as it’s called in Europe); mobile advertising IDs (MAIDs); and the IDs of devices used to consume CTV
  • Dynamic IDs (e.g., ID5’s Universal ID, Panorama, etc.), which start as probabilistic and become increasingly stable as they’re backed by more signals (e.g., hashed email, UID2, several first-party cookies, MAIDs, CTV IDs, etc.)

Together, these identifiers will rebuild the digital advertising landscape for the better. The key will be in ensuring they can operate in harmony with one another.

Connecting the dots of identity at Semasio

The interoperability of the many types of IDs emerging within the market rests on connecting them.  Think of it as a person having multiple passports and exhibiting different behaviors under each of them. The behavior of buying a product could be registered under the UID2, backed by a hashed email. The surfing behavior on media sites could be registered under the ID5 Universal ID and the Panorama ID. App usage could be registered on MAIDs, and CTV viewing behavior could be registered under the CTV ID, the IP address of the households, or even the UID2, as most CTV devices require email authentication.

This interoperability will come from using technology to drive signal recognition across channels. By combining strategies that account for differing types of IDs and their usage, we can ensure the ability to reach a consumer by applying business logic that ensures connections with high levels of fidelity. As an example, a connection between deterministic IDs should be considered stronger and preferable to connections over dynamic IDs, etc.

Does this approach produce the absolute truth about the relationship between users and IDs? Not entirely. But it comes close enough for the purposes of digital marketing and can certainly hold a candle to the third-party cookies we are so reliant on today.

Will this approach work across profiling, targeting, frequency capping, measurement and attribution? It absolutely can if we work together to replace the current cookie-dependent infrastructure with one that is able to recognize users on a number of passports.

The end of the cookie does not have to mean the end of user identifiability.

Go back