Goodbye FLoC, Topics are the new Google targeting system

The experiment of Google FLoC, the system that should have supported online advertising replacing tracking via cookies and guaranteeing the privacy of users, blocked by various disputes, is already over since criticism and lawsuits convinced the American giant to change course: Google has announced that it has diverted efforts on a new project, called Topics, which represents the new attempt to create a new generation of targeting technology.

What are the Topics

Topics is Google’s new proposal to create a mechanism to enable interest-based advertising, without having to resort to tracking sites visited by a user.

As we know, Google has announced its intention to withdraw third-party cookies in Chrome next year and, since 2019, has launched the Privacy Sandbox Initiative to improve web privacy for users, also offering publishers, content creators, publishers (who hosts ads on their website) and other developers the tools they need to create thriving businesses and, at the same time, ensure a safe and healthy Web for everyone.

The fact that “digital advertising is fundamental for many companies” is a way to keep access to online content free, and therefore the interest in the public targeting options that will be available in the near future is particularly high.

Description of Google Topics

It is Vinay Goel – Product Director Privacy Sandbox, Chrome – who officially provides details of Mountain View’s new initiative, which is currently in the initial discussion stage to gather and act on ecosystem feedback and, therefore, it does not yet have a definitive form.

At the moment, however, we know that the Topics API selects the topics of interest based on the user’s browsing history, without involving external servers, and shares them with the participating sites.

In particular, the browser identifies some topics (specifically topics, in English), such as “Fitness” or “Travel”, which “represent the main interests emerged during the navigation of the last week, based on the chronology”. These topics are only kept for three weeks, after which they are automatically removed, and the entire process takes place only on the device used, without involving external servers, even those of Google.

As reported in an article by Search Engine Land, this technology examines “the domain or subdomain of the site to map the topic to that site”, and therefore Google does not analyze the text within an article to determine the appropriate topic. Also, it is made clear that if a site does not participate in the API Topics, “then it does not provide a topic nor receive a topic”.

How the new technology works for ads targeting

In the intentions of the American company there should soon be a trial of Topics for developers in Chrome, and the final iteration of user controls, as well as other technical aspects of the technology’s operation, will be determined based on the test itself and the feedback.

Some more details about this tool can be found in the page created by Sam Dutton – Developer Advocate, Google Chrome – which describes the technical functioning of Topics.

The API Topics proposes a way to provide topics that a user may be currently interested in, based on their recent browsing activity; such topics may complement contextual information to help select the appropriate advertisements.

Basically, it has three main tasks:

• Mapping website host names on topics of interest – for example, a yoga website could be classified as related to “Fitness”.
• Calculate the main topics for a user based on his recent browsing activity.
• Provide a Javascript API to provide topics currently of interest to the user, to allow selection of appropriate ads.

In addition, Dutton continues, Topics can help facilitate robust user controls, as the API is based on recognizable and high-level topics; Chrome plans to offer users the ability to remove individual topics and to show the user the topics they are associated with in the browser, but it should not be possible for users to manually add their own topics.

When a person visits a site that uses Topics, Topics will share with the site and its advertising partners only three topics, one for each of the previous three weeks. Through this system, browsers will offer people significant transparency and control over shared data: for Chrome in particular, the controls should allow users to view the topics identified by Topics and remove those that do not interest, or disable the function completely.

From what is proposed, it is understood that the time factor is crucial: a mechanism to facilitate interest-based advertising, such as Topics, must ensure that the topics of interest provided are kept up to date. According to the initial project, the browser can deduce the topics for a user based on its browsing activity during a period of time known as epochs that, as said, is set per hour to a week: the selected topic for each era is randomly selected from the user’s first five topics for that period of time, and to further enhance privacy and ensure that all topics can be represented, there is a 5% probability that the argument will be randomly selected from all possible arguments in the taxonomy.

For now, Google is starting the initiative with about 300 topics “which represent an intersection between the taxonomy of IAB’s V2 content and also our ad taxonomy review”, but the final number could push up to thousands of interests.

Advantages of Topics for digital advertising

Interest-based advertising (IBA) is a form of personalized advertising in which an ad is selected for a user based on his interests, deducted from the sites he recently visited. This differs from contextual advertising, which aims to match the content on the page that the user is visiting.

Because of its characteristics, IBA can help advertisers reach potential customers and fund websites that otherwise cannot easily monetize visits to their site solely through contextual advertising; In addition, it may also supplement contextual information for the current page to help you find an appropriate advertisement for the visitor.

Google Topics fits into this area, but also offers important reassurances on the privacy side – one of the reasons that led to the premature failure of Floc – because the list of interest-based topics allows companies that advertise online to rely on a useful alternative without opaque tracking techniques (such as browser fingerprinting) to continue to publish relevant ads.

In particular, Google reassures that the set of topics has been carefully taken care not to include potentially sensitive categories, such as gender identities or ethnic connotations. In addition, transparency and control over how data is shared (which is derived from browser history) “will be better and more intuitive than tracking systems such as third-party cookies”.

Differences between Topics and FLoC

The Topics technology now officially appears in the Privacy Sandbox and takes the place of FLoC, definitively withdrawn after a first extension, even in the new project timeline.

The big difference between these two Google targeting proposals is that Topics does not group users into cohorts and therefore limits the risk of using fingerprinting techniques, that could have enabled a user’s browser to be distinguished from the thousands of other users within the same cohort to establish a unique identifier for that browser.

In addition, in the FLoC proposal presented and tested by Google, the browser collected data on a user’s browsing habits to assign that user to a cohort, with new cohorts assigned on a weekly basis, based on the browsing data of the previous week. The Topics API, on the other hand, determines which topics to associate to the user on a weekly basis based on the browsing history, but these topics are kept for three weeks.

Technically, then, Floc shares a cohort ID, while Topics selects three topics to share with sites and advertisers.

It is not wrong to think and argue that, basically, Topics arises from negative criticism from many sides against Floc, which since the origin tests in 2021 has received a wide range of feedback from contributors to the web ecosystem and adtech. In particular, there were concerns that Floc cohorts could be used as a fingerprint detection surface to identify users or that they could reveal the association of a user with a sensitive category, and there were also appeals to make Floc more transparent and understandable for users.

Topics, a safer targeting system?

With the new system, however, Google has tried to provide a different way to support interest-based advertising, with greater transparency, stronger privacy guarantees and a different approach for sensitive categories.

In detail, it proposes multiple mechanisms to ensure that it is difficult to re-identify a significant number of users on sites using only the Topics API:

• The taxonomy of the arguments provides a series of coarse topics (for now there are about 350 in total, as said), and therefore it is likely that each argument has a large number of users and there is a guaranteed minimum number of users per topic.
• The arguments are returned randomly by the first five of the user.
• 5% of the time a random topic is provided (chosen from the complete set of topics).
• If a user often visits the same site (for example, every week), the code running on the site can learn at most one new topic per week.
• Different sites will receive separate topics for the same user in the same era. To make it more difficult to determine whether the user is the same, there is only one in five possibility that the topic returned to one user on one site corresponds to the topic returned to them on another.
• The topics are updated to a user once a week, which limits the speed at which the information can be shared.
• A topic will only be returned to an API caller who previously observed the same topic for the same user recently. This model helps limit the potential for entities to know (or share) information about the interests of users who have not personally observed.

Goodbye to Google FLoC, a controversial project

The new approach of transparency towards the user chosen by Google is already in itself a step forward compared to the great weaknesses of Floc, in particular the Ids of cohorts.

In its original form, the FLoC or Federated Learning of Cohorts technology provided for the division of Internet users into cohorts on the basis of the pages visited and searches carried out; in this way, companies could advertise interestedbased not on the individual, but on a “coarse” group of users.

A kind of compromise that goes beyond cookies – which are able to accurately monitor the preferences and behavior of individual users and that help publishers who want to sell their advertising space in a targeted way – with the attempt to dilute the identity of the single user through the widening of the monitoring to the monitoring of a sample “packet” of users, that however has been strongly criticized because it did not completely assure the protection of the privacy of the users (and because, no less important, it left vertical profiling in the hands of Google).

The attack from german publishers

With regard to criticisms and attacks, for example, only a few days ago the press reported on the initiative of the German publishers, who presented a formal request to the European Competition Commission to protest against Google’s policy on cookies.

In particular, two advertising companies and some of the main German publishers believe that the unilateral abandonment of cookies will cause a significant impact on the advertising publishing market which will benefit Google itself, that through Chrome owns a major share of the online advertising market, amounting to 60%, while (according to the estimates presented) because of the cookie apocalypse publishing companies are expected to lose a percentage of revenue ranging from 20 to 70 percent.

What is the future for the online adv

In short, the future of online adv and a fundamental theme also for the information market and the production of content on the Web is at a turning point.

Google’s first step, Floc, has run aground due to rejection by industry operators, regulatory problems and lack of adoption, and the real challenge for Topics will be to convince other big players to use the system.

At the moment, Google is in the early stages of implementing the API Topics, so other browsers probably have not yet had the opportunity to assess the technology well; based on the poor reception of FLoC, however, you can imagine that even in this case Firefox, Safari, Edge or other browsers refuse to adopt Google’s proposal, and this of course may affect the size of the user base that advertisers can tap into.

Only time will tell what will be the fate of Topics and targeting for advertising purposes, and on this aspect for now Google continues to keep the timeline already set, which plans to make available for adoption all Privacy Sandbox APIs during the fourth quarter of 2022, although of course there may be further news.