Google Shifts to Topic Modeling

Google already committed to eventually doing away with third-party cookies. As a substitute, it last year introduced Federated Learning of Cohorts, or FLoC, for interest-based targeting. FLoC, as it was proposed, could bucket users into cohorts each week based on their browsing histories then report those anonymous cohorts back to advertisers for targeting.

FLoC had flaws: The concept was hard for consumers to grasp, and it raised a lot of privacy and security questions and concerns, so Google in January ended FLoC and replaced it with Topics API, another contextual targeting solution powered by consumer browsing history.

The main difference between FLoC and Topics API is that, instead of assigning users cohort IDs every week, it will assign consumers three to five topics of interest every three weeks based on their browsing histories. The Topics API currently contains 350 topics, but that number could grow significantly. The API will then report three of those topics to participating sites, which they can then use for advertising.

Another major difference is that with Topics, sites can opt out of being assigned a topic and users can see the topics to which they’ve been linked.

But critics suggest that Topics might not be the answer either. Among their concerns with Topics, the data it would provide, they say, is less precise than FLoC.

“We know that content alone does not provide the level of accuracy that the combination of contextual and real-time data signals offers,” says Yahoo’s chief business officer, Ivan Markman. “The industry will need greater granularity and alignment with user relevance.”

Still, Markman, like many others, isn’t surprised by the Google move.

“The evolution of FLoC to Topics API is indicative of the challenge the industry is facing on identity: respect consumer preferences and create a better value exchange, while mitigating and maintaining advertiser and publisher relevance and reach,” he says.

Krystal Tang, senior manager of solutions at Uberall, shares a similar notion.

“With the release of Topics, Google is trying to deliver more unique and personalized experiences in a way that users opt in to. The move by Google is likely related to the company’s broader initiative to phase out third-party cookies, finding new solutions that meet the needs of advertisers while still protecting user privacy. It will be interesting to see how this is adapted overtime,” she says.

Evan Hills, senior vice president of partnerships and strategy at Dstillery, has his doubts. “Google is suggesting all third parties replace their unique targeting methods with a single shared method. Even if developed with industry participation, it replaces the current diverse ecosystem of unique value propositions with a single commodity offering, effectively precluding competition,” he says.

“The industry needs a better understanding of the consequences of the Topics model, given how quickly the cookie deprecation is coming,” Hills adds.