Google’s quest to find a way forward for targeted advertising in a more privacy-centric world continues unabashed, despite its many (and vocal) detractors. After its previous proposal to replace the third-party cookie’s ad targeting capabilities came under fire for being potentially too invasive, Google has introduced an alternative cookie alternative called “Topics” that swings the pendulum in the other direction.
WTF is Google’s Topics?
Topics is among Google’s attempts to replace the third-party cookie as a means of identifying people online by striking a balance between preserving people’s privacy and preserving companies’ abilities to buy and sell targeted ads.
Third-party cookies have been the connective tissue for the online advertising ecosystem since its very inception in the 1990s, but Google has decided to disable third-party cookies in its popular Chrome browser in 2023. However, establishing consensus on how the industry can move ahead has proven tricky.
Wait, haven’t they tried this before with all the bird-themed stuff we heard about last year?
Yes, kind of, but it’s probably best that we provide some context here.
The simple fact is that Google’s previous cookie-replacement proposal — Federated Learning of Cohorts — really did put the cat amongst the pigeons if we are to stick with the aviary-themed wording. Why? Simply put, the concept of FLoCs was laden with privacy concerns (ironic, no?) and didn’t even pass muster under laws such as the European Union’s privacy law, the GDPR.
Rather than the cookie-based approach of targeting people at the individual level based on their browsing behaviors, the FLoC approach would clump people into specific interest groups — cohorts — such as people in the market to buy a car or impulse shoppers, based on their browsing history. These clusters were supposed to provide cover and give people a greater degree of anonymity online. But that privacy shield was found to be permeable.
For instance, during a trial phase, researchers raised concerns that FLoC data could be combined with people’s personally identifiable information. This meant that bad actors could expose information about people’s webpage visits and interests, raising further concerns that cohort-based targeting could be used to discriminate against particular groups of people.
However, what FLoC did introduce was the concept of using machine learning to categorize users into specific interest groups based on their browsing history. Also importantly, FLoC proposed performing this categorization on a person’s computer or phone, rather than sending the underlying data to Google’s or another company’s computers, which went some way to better preserving people’s information.
OK, Google Topics, how does it work, and how is it different from before?
So, after the FLoC blowback, Google went back to the drawing board and came up with Topics. Like FLoC, Topics proposes targeting ads to people based on the categories of content they check out online and containing this categorization to the device. Unlike FLoC, though, Topics proposes a much more general level of categorization.
How does Topics work?
A web browser like Google’s Chrome will use the Topics API — short for “application programming interface” and effectively the means of using the Topics toolset — to determine a number of topics that best reflect a person’s interests based on their browsing behavior, such as “autos & vehicles,” “basketball,” “news,” and “women’s clothing.” The browser will determine the topic of sites based on sites’ hostname — ex. dogs.com would be categorized under the “dogs” topic — but companies may be given the option of declaring which topics to associate with their sites.
Each week, a browser would select five topics per person — including one random topic that is meant to throw off any companies attempting to wrest a person’s identity from Topics — and then choose one topic, or interest category, to assign to that person for the week. People may be given the option of adjusting the topics assigned to them and will also be able to disable this ad targeting feature.
Each topic is then kept for three weeks so when a person visits a website, the website or the ad tech firms it uses to target ads can use the Topics API to access up to three topics for that site visitor. However, the website and its ad tech firms can only access topics that are related to the given website or other websites carrying the ad tech firms’ code.
In other words, if a person assigned the topics “dogs,” “theme parks” and “weddings” visits a political news site, that site may not be able to access any topics for that person that week. But if that person is assigned the topic “politics” the following week and returns to the site, then the site would be able to access the “politics” topic at the least.
How many topics are there?
Currently, there are 350 interest groups classified by the Topics API. Compare this with the 30,000-plus user classification groups that featured in the FLoC trials. Critics said the granularity of the FLoC API made it possible to reverse-engineer data and pinpoint user data. The initial 350 topics are only a starting point, though. The list may eventually number into the thousands, though Google plans to come up with a number of sensitive topics to exclude from the list and to have an outside party be responsible for providing the list of topics to be included.
What has the reaction been so far?
Mostly, it’s been speculative. Trials have not yet begun, and Google hasn’t publicly said when it will start letting third-party developers experiment with the Topics API.
However, advertisers and media owners alike have voiced concerns that Topics’ lack of granularity will limit the specificity of ad targeting and, as a result, limit or even lower ad prices, with advertisers typically paying more money for more targeted ads.
Ad tech execs who work with publishers have also asked for greater clarity from Google over the extent to which they will be able to control the flow of their user data should they integrate the Topics API. One concern is whether a third-party website could capitalize on the data extracted from a publisher’s website via the API to boost the third-party site’s own ad sales. And, after all, there is the potential for that third-party website to be a direct competitor.
How do I prepare?
For advertisers and publishers, the idea is still the same as it was in January 2020 when Google Chrome confirmed plans to withdraw support for third-party cookies: collect as much consented first-party user data as possible.
Publishers with a glass-half-full outlook on Chrome’s deprecation of third-party cookies view the move as offering the potential for them to charge advertisers more money for access to their first-party data. Meanwhile, advertisers are under pressure to adapt to a post-cookie landscape or risk missing out on reaching potential customers online. For ad tech companies, arguably those who are most exposed to the whims of the Google Chrome team, the best possible advice is to roll with the punches, be agile, and prepare to pivot if necessary.
Ultimately, every tier of the online advertising industry needs to participate with Google as it prepares to forge a path forward. Simply pointing the finger and complaining over Google’s dominance of the sector is unlikely to yield dividends any time soon.