When Google Chrome delayed the deprecation of third-party cookies last year, it signalled that big changes were imminent. Sure enough, on January 25, Google announced the Topics API, a new Privacy Sandbox proposal for interest-based advertising. Topics replaces the Federated Learning of Cohorts (FLoC) proposal and will reflect a consumer’s interests, rather than placing them into interest-based cohorts like FLoC.
To those who’ve been following the evolution of the Privacy Sandbox proposals, this isn’t a surprise. Concerns about FLoC’s compatibility with GDPR, and its ability to protect consumer privacy while offering transparency into data use, led Google to reflect and develop this new proposal.
The Topics API will continue to evolve as Google engages the open ad tech community, including Index, for feedback. However, compared to the prior proposal, Topics appears to offer a more intuitive and privacy-centric way to provide addressability in Chrome after third-party cookie deprecation.
What is the Topics API?
With Topics, consumers using the Chrome browser are grouped by topical interests based on the content they view. For example, a consumer reading articles about the 2022 Olympics might be added to a “Sports” interest group. When they eventually visit another website that uses Topics, the API sends up to three topical interest labels to that publisher’s advertising partners, signalling which ad categories might be of interest to that consumer.
The API’s decentralised approach to consumer data should please many privacy advocates. Topics are stored on the consumer’s browser, not an external server belonging to Google or another party. According to Google, “Topics are kept for only three weeks and old topics are deleted.” A further indication of increased consumer privacy is that the Topics taxonomy doesn’t include sensitive information such as ethnicity or gender.
What are the improvements over FLoC?
- Better privacy: Advocates for consumer privacy were concerned that FLoC’s design allowed consumers to be tracked based on sensitive information. While FLoC calculations excluded content within sensitive categories, it was still possible to infer sensitive topics from FLoC clusters, especially with more advanced machine learning techniques. There was also uncertainty over how third parties might use FLoC IDs.
Topics improves upon this by grouping consumers explicitly by topics of interest, and communicating the interest itself rather than the group identifier. There are also fewer, more generalised groups: FLoC had about 30k cohorts, while the Topics API GitHub page only lists about 350 category labels.
- Improved transparency and control: FLoC gave consumers limited insight into how their data would be interpreted. It was difficult (by design) to understand what interests a cohort shared. This made it hard for media owners, too, as they’d have to rely on partners or develop their own insights about the correlation of cohort IDs to different types of content.
With Topics, consumers will be able to see which groups they belong to, remove themselves from groups, or opt out entirely.
- More turnkey and intuitive: FLoC IDs had no inherent meaning—not even Chrome knew why certain consumers were grouped together (that’s what the “federated” part of the name meant). Parties outside of Google would have to conduct additional intelligence and data science to make those IDs actionable for marketers. Topics’ plain language taxonomy provides an off-the-shelf solution with interest-based audiences that are easy to understand. This makes it easier for media owners of all sizes to get aggregated insights about their consumers’ high-level interests.
You can read more about the changes and evolution from FLoC on Google’s GitHub page.
What are the ongoing areas for discussion?
Despite the improvements over FLoC, there are still a few concerns and open questions from privacy advocates and marketers. First, the Topics API still fundamentally exposes consumer interests for advertising purposes, even if it doesn’t expose the consumer. For some privacy advocates, this will continue to be unacceptable when compared to, for example, purely contextual targeting solutions.
Second, marketers are keen to understand the effectiveness of the Topics API. Topics are much less granular than FLoC cohort IDs were, so marketers might discover that they’re too generalised for impactful ad targeting.
What are the implications for media owners and buyers?
The post-cookie web will be divided into two groups: authenticated and non-authenticated audiences. The challenge for marketers is how to reach these audiences; the challenge for publishers is how to monetise them. The Topics API seems to offer an intuitive and performant way to message and monetise non-authenticated audiences using the massive reach of Chrome, which accounts for almost 65% of global market share.
We’ll understand more as the industry puts the Topics API through its paces. However, the first look is promising.
The open web relies on marketing investment and value exchange. In return, we’re able to enjoy amazing content that enriches our lives. The Topics API proposal lays the groundwork for the continued health of a vibrant open web after third-party cookies go away.
