FEW PEOPLE outside the advertising industry will mourn the death of third-party cookies. These are snippets of code placed on people’s browsers that allow advertising companies to track them around the web, gather information about them and then flog products to them based on their interests. (They differ from first-party cookies, which websites use to remember people’s details, for example to keep them logged in, but do not track them when they leave.) Last year Google announced that by 2022 it would no longer use third-party cookies or permit other companies to deploy them via Google Chrome, the internet’s most popular web browser. Other big browsers have already blocked third-party cookies by default. The technology seems moribund, but Google has just started testing what it hopes will become the new standard, called Federated Learning of Cohorts (FLoC). It has already been criticised by the makers of other browsers and the Electronic Frontier Foundation, a digital-rights group. What is FLoC, and why is it controversial?
Instead of allowing advertisers to track individuals’ browsing history, as with cookies, FLoC groups similar browsing behaviour into numbered “cohorts”, each of which contains thousands of users. Each week a person’s browser will review the sites they have visited and match them to a cohort. Advertisers can see which cohort a person belongs to, but not the characteristics that link its members. But by observing the behaviour of other members of this cohort, advertisers can work out the interests they probably share. If, for example, lots of members of one cohort search for water bottles, advertisers might serve other members ads for bottles, too. Google claims that this semi-anonymity is more protective of privacy than third-party cookies, which track specific users, but still 95% as effective at getting users to click on ads. But FLoC is coming under fire.
Many people will still balk at the idea of a label that can be used to infer their interests, even if advertisers aren’t tracking their every move online. Although users can block FLoC, if it is enabled by default, as it probably will be in Chrome, many will not. It may also be difficult for Google to stop the system from grouping people by characteristics they wish to keep private, such as race or sexuality. Beyond this, the system may not be as anonymous as it seems. Some critics worry that sorting users into cohorts will make them easier to “fingerprint”, whereby websites use other identifiers, such as your IP address or even the size of your browser window, to identify individual users. FLoC would give them one more way to distinguish people.
FLoC has also been accused of being anti-competitive. Companies that use it would be reliant on semi-anonymised data to identify people. But Google, one of the biggest players in online advertising, is not so hobbled. It amasses lots of first-party data in other ways, such as when users interact with Chrome or Android, its mobile operating system. This gives Google a competitive advantage.
Google began testing FLoC in March by rolling it out to 0.5% of Chrome users in ten countries. No other browser has said it will implement FLoC, but two, Brave and Vivaldi, have come out against it, as has DuckDuckGo, a search engine. Microsoft has disabled FLoC in Edge, its browser. Even if Google tweaks FLoC in light of the trial it will probably still be a hard sell. But that might matter little. Given Google’s dominance of online advertising and the ubiquitousness of Chrome and Android, FLoC might steamroller its way to being part of the architecture of the web.
Dig deeper:
A new type of ad is heading for your iPhone
The advertising business is becoming less cyclical—and more concentrated
Google’s problems are bigger than just the antitrust case