Topics to Replace Cohorts: How Do Businesses React to Topics API

Despite the undeniable dependence on third-party cookies the current iteration of digital advertising technology has, Big Tech will soon abandon this model and advance to something else. Most modern web browsers either enable users to switch off 3rd-party tracking cookies or have that feature set by default.

First, Mozilla’s Firefox implemented this approach, then Apple’s Safari, and this year – the biggest browser of all – Google Chrome. The latter is particularly significant as Chrome has up to 65% of the web browsing market as of January 2022, according to W3Counter.

Google’s move was inspired by the global quest to improve digital privacy and stimulated by governmental authorities as well as public opinion. But before Google can completely give up on the effective and familiar technology, the company has to come up with a replacement.

Google is trying to come up with a stable and privacy-secure solution that would not damage its ad targeting precision. Their first technology, called FLoC, promised to solve the controversy but failed to convince the community of its effectiveness. The second big proposition from Google named Topics API currently struggles for the same reason. Or is it? Will businesses adjust to it? Let’s find out.

Was Google FLoC Really a Misstep?

FLoC (Federated Learning of Cohorts), which is often called the “future of internet advertising”, is based on Privacy Sandbox. Despite the ‘federated learning’ part in the name of the technology, it did not actually utilize any federated learning technologies, which implies using machine learning algorithms. The main difference from cookie technology is the availability of nondescript anonymized cohorts based on the users’ common interests.

The idea of marking groups based on interests was met with opposition in the web advertising industry. Google started the testing in March 2021 and then canceled it in July 2021. During the trial, Google received mostly negative feedback from the companies that participated, making it impossible to encourage the industry to use the innovation. It’s also illustrative how all other Chromium-based browsers dodged implementing FLoC functionality in their browsers while other tech companies like Amazon blocked it by default.

The project was silently removed from Google’s roadmap without further explanation, and with no intention to share testing insights. Some of the known concerns associated with FLoC were tied to the possibility of identifying users through fingerprinting, the system’s over-complexity, and lack of transparency.

But despite the technology failing to find more followers, its testing can be considered successful. Google gathered all the feedback they needed to develop a better solution. This is evidenced by the decision not to extend the test, although there were many companies still willing to contribute.

Are Topics API Going to Save the Party?

As the FLoC experiment ended, Google utilized the feedback from the community to integrate it with Topics API. Like its predecessor, the system was built on interest-based advertising, only this time, it would distinguish five main interest categories from browsing activity in recent weeks and not use any external data. The API in the name infers user categories on a website via a classifier model that maps hostnames to topics that are publicly available with Chrome Developer Tools.

The algorithm does not analyze additional information like full URLs or website content to improve privacy and proposes mechanisms to opt-out of API for both users and websites. Upon visiting a website, Topics API will notify publishers of three categories of interest, one for each of the past weeks in random order.

The advertising taxonomy is designed to include between a few hundred to a few thousand topics and currently contains around 350 categories like “travel”, “cooking”, or “gaming”.

That sounds great on paper, but the reality is more prosaic. Practically the same issues that deterred Cohorts from gaining momentum are now building obstacles for Topics. Thus, the ethical dilemma remains: FLoC technology was meant to share personal data to a wide range of publishers with no direct approval from users, while Topics API features several security ‘upgrades’ that only minimize the probability of tracking but do not eliminate it entirely.

Despite this fact, Topics’ targeting capabilities still remain poor, which is largely acceptable for bigger brands with broad audiences of the likes of Coca-Cola but fatal for small and medium advertisers. The categories-based solution is simply too imprecise to satisfy narrow targeting criteria and campaign budgets – that’s why smaller businesses should probably opt for alternatives.

Are There Alternatives to Topics API?

Many medium-sized businesses that, until today, were satisfied to work with Google, now will likely face hindrances in reaching narrow audiences. This pushes them to look for substitutes, at least temporarily, until big tech provides a working solution. One such alternative can be found via purchasing ready-made audience segments from trustworthy data providers to run campaigns on their channels.

Buying audience data from data brokers or specialized platforms has great potential to solve the precision issue through data-driven advertising. The technology essentially utilizes targeting based on first-party user data that conforms to applicable laws and self-regulatory standards, meaning it’s not severely limited by privacy regulations.

What it does is it segments user data into cohorts that feature similar characteristics such as purchase intentions, demographic, and interests to offer advertisers ad spots for reaching those people.

Here are some examples from our practice that helped solve the problem with a narrow audience:

1. A pharmaceutical company focused on the prevention of certain narrow diseases. Medicine cannot provide the same solution for all problems, so we tried to segment the audience as accurately as possible. We have been able to build an audience through data partners that meet specific age and medical requirements. For example, aged 55+ with high blood pressure. Let's clarify that the campaign was promoted in countries and on sites where there are no restrictions on such advertising.

   
   

2. You might think that everyone buys jewelry, but it all depends on the specifics of the products. Therefore, a wide audience did not bring such a result. We built an audience of males who have been married for 1-3 years. Data from our internal research showed that this audience is more likely to spend on jewelry for certain events (anniversaries, birthdays, etc.).

   
   

3. The purpose of advertising is often to recruit staff. We had a goal to find suitable candidates for the summer camp. We have created a dedicated audience for college students and job seekers. Our client reached their HR target because of the proper targeting we provided them with.

What Lies Ahead

Google’s attempt to swap 3rd-party cookies for FLoC wasn’t successful. As for Topic API technology, the ad tech industry expresses a glimmer of hope that it will bring more value than its predecessor.

Browsing history-based targeting is a more secure way of delivering ads compared to the old-school cookie packages and an obvious improvement over FLoC, but it still has miles to go before we can call it a privacy-respecting marketing tool or an effective targeting tool suitable for smaller advertisers.

Big tech is currently struggling with privacy-related regulations, and I believe the industry will adapt to further restrictions through decentralization and transition to first-party data advertising.