1st April 2021

POV: Google’s Federated Learning of Cohorts (FLoC)


Google has started to roll out trials of Federated Learning of Cohorts (FLoC), the Google preferred alternative to third party cookies that enables interest-based advertising.

Details and Implications:

The trial will allow websites to begin testing without asking browser users to turn on specific flags and will start in the US, Australia, Brazil, Canada, India, Indonesia, Japan, Mexico, New Zealand and the Philippines but will be scaled globally over time. Users will be able to opt out of the trial should they wish to. At this time, FLoC is still under development and Google says that its aims to learn from these trials and evolve it accordingly.

Third-party cookies are expected to be blocked on Chrome by next year (and already are blocked on Safari and Firefox) and so, although cookies won’t be replaced like for like, alternatives are being explored in order to allow targeted advertising to continue. The FLoC trial is one of multiple privacy sandbox initiatives being explored by Google to achieve this.

A major issue with the cookie is that it can personally identify users and therefore poses data privacy concerns as more stringent data privacy laws such as the GDPR are instituted by governments around the word. FLoC runs locally on your browser and instead of a personally identifiable cookie, analyses your browsing behaviour in order to group you with other people with similar interests.

According to the proposed approach, the Chrome browser will allocate users to groups (called cohorts) based on their browsing behaviour. Advertising is then targeted based on membership of those cohorts.

For example, if you visit a website about ‘golf’, a cohort ID will be stored on your browser to say that you are interested in this subject and then you may see relevant golf-related ads when you visit other websites if they are using advertising technology that uses the FLoC cohort IDs. All the browser displays to the ad technology is the cohort ID, with your data and browsing history staying on your local device. FLoC also uses machine learning models to predict the probability that a user will convert, based on the cohort, and can also recommend content to users.

The aim of the cohorts approach is to allow advertisers to show relevant ads to users without the fears over personally identifying individuals, meaning consumers can continue to see relevant ads.

Also this week an alternative to FLoC called ‘SWAN’ was revealed, supported by a range of ad tech companies including PubMatic, OpenX and Zeta Global. There is now a 60-day period to allow trials of SWAN before a planned summer launch and the initiative is intended to be overseen by a third party organization, much like domain names are overseen by ICANN . A further solution, Unified ID 2.0, is also being backed by TradeDesk, which plans to pass oversight of the project to non-profit industry associations such as Prebid.


It will be interesting to see how the FLoC trial develops overtime and how effective it is for advertisers going forward into a cookie-less world.

Further Reading

TechCrunch | The Verge | web.dev

Mindshare Global
    Mindshare Global