Ultimate Guide to Federated Learning of Cohorts (FLoC) in 2024

As third party cookies are leaving our life, Google offers marketers Federated Learning of Cohorts (FLoC) so they can keep optimizing their ad spend.

FLoC is a new tracking technology that is planned to be rolled out as third party cookies are banned by Google in 2023. FLoC uses federated learning principles. FLoC creates groups or “cohorts” by using browser historical data of users. Google has just started to test FLoC in 2021 in some countries.

What is Federated Learning of Cohorts (FLoC)?

Google intends to replace the third party cookies with FLoC. FLoC is a browser-based model that can track the users’ behavior on websites and then combine this information into cohorts of interests. Marketers will be able to use these cohorts for targeted advertising. Collected data via FLoC will not sent to external servers, but the browser will have control on the digital advertisement.

FLoC will be used for showing relevant ads to the cohorts. As a system proposed by Google, it passes power from media companies to Google. Google dominates the browser marketplace with Chrome and in this model a system in Chrome built by Google will be playing an active role in choosing which ads to show the user. Historically, it was the media companies or ad exchanges that decided which ad would be shown to a user.

How does FLoC work?

FLoC algorithm will

• not use individual ID data, use browsing history. The data collected from the browser will be saved locally on the visitor’s computer. This is a privacy-preserving feature.
• generate anonymous groups which are called the cohorts in which visitors have similar interests.
• assign cohort IDs to each cohort and update the cohorts weekly
• communicate to the website which cohort the visitor belongs every time the visitor visits the website. This means that people with similar interests will see similar ads.

There are rules to track and use a cohort for targeting ads. For example a cohort must have a minimum of 1,000 users.  Advertisers will not see any personal data about an individual, they will access common information about the users which are in the same cohort.

The following image is an example to explain different roles, when using a FLoC algorithm.

What is the level of interest in FLoC?

FLoC has started to be popular after Google announced plans of rolling it out. While FLoC is being tested, companies especially adtech companies and privacy advocates are exploring its impact. The interest in FLoC is seen below:

What is the next ?

Impact: Google is still continuing to develop FLoC solution, so it is not possible to say something about its effectiveness certainly in advance.

Alignment with data privacy regulations: How FLoC solution will comply with the The General Data Protection Regulation (GDPR) is an issue pending clarification. Under GDPR, creating the cohorts from users’ data can be interpreted as use of personal information. Google announced that it would not run initial trials of FLoC in EU countries, due to GDPR. FLoC testing is running in the US, Mexico, Canada, Australia, New Zealand, Brazil, India, Indonesia, Japan, the Philippines.

Other concerns: There are concerns about combining “digital fingerprinting” with FLoC to de-anonymize FLoC’s anonymous groups and use FLoC to monitor users with even a higher level of precision compared to the past.

For more, feel free to read our articles about:

• First Party Data
• Top 10 Privacy Enhancing Technologies (PETs) in 2021
• Differential Privacy
• Third party cookies ban

If you have questions about FLoC, we would like to help:

This article was drafted by former AIMultiple industry analyst Ayşegül Takımoğlu.