Topics API for Privacy Sandbox: All you need to know

On January 25, 2022, Google announced Topics, a new Privacy Sandbox proposal that will replace their previous one, called Federated Learning of Cohorts (FLoC). Both FLoC and Topics have to do with increasing privacy of individuals and Google’s plans to stop using third-party cookies on Chrome. The cookiepocalypse, as the term has been coined, has been delayed for 2023, and Google is trying to find a solution that will still allow for online advertising to continue thriving without the invasion of user privacy.

What is Topics API?

Topics API is meant to create grounds for interest-based advertising. The browser will collect information about the user’s top five topics of interest on a weekly basis based on his or her browsing history. The topics will be stored for three weeks on-device and then deleted from the browser. The Topics API will share only one topic from each of the previous three weeks (a total of three) with the sites you visit and their respective advertising partners. Chrome users will be able to see which topics are selected for them, delete any they don’t find relevant, or disable the functionality entirely. Google has reported that Topics API will be human-curated to not include any sensitive topics, such as gender, race, and religion. 

How does it work exactly?

Each website will get a high-level topic “label” from the API, selected from a human-curated list of topics. Currently, the proposed list contains about 350 topics, a combination between Google’s own taxonomy and the IAB Tech Lab taxonomy. When the user visits websites over the course of a week, the browser will collect up to five of the most frequent topics that are associated with the user’s online activity. The API will assign a sixth topic at random, which can be sent to advertisers in 5% of the cases, to ensure each topic has a minimum number of members and provide a form of plausible deniability. From then on, every week the API will share one topic from each of the previous three weeks with the sites the user visits. This will help advertisers show relevant ads without having to know the specific sites the user visited. After three weeks the topics are deleted from the browser and don’t go through any external servers, including Google’s. 

Publishers will decide whether or not to include the code that calls the Topics API and if they decide not to, the respective site will not be included in the browsing history eligible for frequency calculations, which is basically a site-level opt-out.

Users, too, can opt-out from specific topics, or entirely from any records about topics through Chrome’s additional functionalities that are currently being built. 

How is Topics API different from FLoC?

Both FLoC and Topics track user interests and report them to advertisers in order to increase ad relevance and, respectively, ad clicks. However, they do so in a different way and allow for different user controls. FLoC groups users in cohorts based on their behavior, while Topics uses interest topics as the main point of reference. Furthermore, with Topics, users will have more transparency and be less individually identifiable. On Chrome, they will also be able to remove certain topics they don’t find relevant for them or simply don’t want to see ads about, or they can entirely opt-out from this feature.

FLoC was designed to generate a cluster ID when a user visits a website, which allows for building users’ browsing patterns and grouping them together with users with similar patterns. However, there were concerns that these groups would be so small that users might still be identifiable. With Topics, advertisers will only receive three topics of user interest, one from each of the previous three weeks of their online activity across participating sites (those that choose to call the Topics API). These topics should remain broad enough so that users are not individually tracked and are deleted after three weeks. For comparison, Topics currently has around 350 topics of interest that can be assigned to a website and this taxonomy is expected to be expanded but still stay in the hundreds or lower thousands, while FLoC has about 32+K. The restricted, human-curated list of topics in the newly proposed privacy solution is expected to further reduce browser fingerprinting and will be free of sensitive topics, such as race, sexual orientation, and religion.

One of the main criticisms that FLoC received was that it is not really a privacy protection advance when compared to third-party cookies, which allow advertisers to track users and monitor their individual browsing history. The newly proposed Topics API provides advertisers with much fewer data and this data expires much more quickly, which is a significant improvement from both third-party cookies and FLoC. It is worth noting that other browser vendors, such as Mozilla (back in 2019) and Apple (2020), have decided to completely block third-party cookies in the interest of user privacy. Google, however, is still postponing this decision (currently pushed back for 2023) and looking for alternative approaches as 80% of their revenues come from advertising. 

What’s next?

Google expects to make Topics API available for developer testing in Chrome toward the end of Q1 2022. The trial will allow brands and publishers to test the technical aspects of the tool and provide feedback on user controls, design, and overall usability. Adoption by other browsers remains unlikely despite addressing many of the issues with FLoC, but only time will tell how big of a user base advertisers will be able to work with post cookiepocalypse.