In the latest move in the privacy wars, Google has announced they’re launching Topics as an alternative for their much-maligned plan to replace third-party cookies with the Federated Learning of Cohorts (FLoC). Where FLoCs proposed using browser data to assign a user to a “cohort”, Topics will instead leverage browsing history to represent a user’s top five weekly interests and only store data for three weeks per browser. In theory, this would alleviate many privacy concerns associated with techniques such as fingerprinting where a specific user could still be personally identified. Additionally, Topic selection allegedly does not involve any external servers, including Google’s. Using the IAB’s Content Taxonomy and Google’s own learnings, there will be around 300 potential Topics for advertisers to target at first—potentially scaling much higher.
As a media strategist and buyer, my expectation is that buying capabilities will evolve from typical contextual targeting (where we select the content where ads will appear) to what will essentially be contextual retargeting. In essence, contextual 2.0. It should be clearly stated that this is NOT a new tactic by any stretch, but rather a canned capability that begins to standardize the practice of retargeting audiences based on browsing habits (not a bad thing on Google’s part).
Here’s where things get a little grey. Sites have to opt-in to participate, and it seems ambiguous at this time how search data may play into the assignment of Topics. Plus, with other browsers likely not buying into Google’s technology, this still doesn’t alleviate the challenge of targeting across browsers or devices—something that’s key to limiting wasteful spending on the part of advertisers.
With FLoC now behind us and Google officially moving to Topics API as the next trial for their Privacy Sandbox, we’ll be keeping a very close eye on how these capabilities evolve and test when appropriate. The ability to scale ad buys will be key to successful deployment, and only time will tell if publisher buy-in will allow for this.