Goodbye Google FLoC, Hello Topics and FLEDGE

Dionysia Kontotasiou
February 23, 2022 ·
Goodbye Google FLoC, Hello Topics and FLEDGE

What is FloC?

In August 2019, Google announced a new initiative, the Privacy Sandbox, to make the web more private for users. And the heart of that initiative carried the idea of killing third-party cookies. Many marketers feared that this could destroy online advertising, or at least the present-day advertising.

Privacy Sandbox proposals
Source: Privacy Sandbox proposals’ timeline

Google submitted these ideas to the W3C, the web’s standards-setting body. After the submission, the proposals were discussed in the web advertising business group, an organization that mainly constitutes ad-tech vendors.

Google and other advertisers (Criteo, Magnate, NextRoll) proposed hundreds of technical standards. Each of the suggestions is intended to take over one of the jobs that cookies now do in the targeted advertising ecosystem.

That’s how FLoC was created.

Federated Learning of Cohorts (FLoC) has a simple plan – to replace cookies with interest-based advertising.

A FLoC compliant browser starts to collect data about users’ interests and then uses that data to assign the users into “cohorts.” For example, a cohort could have an “interest in gardening.”

In short, this is how FLoC works:

  • Users with similar browsing habits are grouped together.
  • Websites and advertisers share a cohort ID with each user’s browser, identifying the group they belong to.
  • Each cohort needs to have at least a few thousand users. If too small, it’s coupled with other cohorts until it reaches a reasonable size.

Sounds complicated?

Look at it this way: your FLoC ID is a concise description of your recent online activities.

how Federated Learning of Cohorts (FLoC) works
Source: Google’s infographic on how FLoC works.

The domains of the websites visited by each user are used as the basis for bringing people together in Google’s proof of concept. After this, Google uses an algorithm called SimHash, to create cohorts. Since SimHash could be computed locally on each user’s computer, there’s no need for a central server to collect behavioral data. However, a central administrator may have a role in ensuring privacy guarantees.

To avoid any cohorts from being too small, Google opted for a strategy that involved a central actor counting the number of users assigned to each cohort. So if any of the cohorts are too small, they will be merged with a similar cohort until each has a sufficient number of users.

How to Know if you’re FLoC’ed

When Chrome 89 was launched in March 2021, it included a FLoC test run and you may have already been enrolled without realizing it.

To find out if you’re FLoCed, go to EFF’s Am I FLoCed? webpage.

If you don’t want to be a part of the trial, go to the Chrome Settings window and select “Privacy and security”, then “Cookies and other site data”, and uncheck the “Block third-party cookies” option.

The Backlash against FLoC

When Google rolled out the beta version of FLoC, it was met with opposition and distrust from advertisers and privacy groups alike.

The EFF was one of the first to speak out against Google’s idea, when it wrote “Google’s FLoC is a Terrible Idea”.

FLoC has also been condemned by other major browsers, like Mozilla:

We are presently considering many of the privacy-preserving advertising approaches, including some proposed by Google, but have no plans to implement any of them at this time.

Safari did not release any official statement, but Webkit Engineer John Wilander had published this:

Brave has a lengthy essay explaining why it has disabled FLoC, claiming that it is causing harm to users and is a step in the wrong direction.

DuckDuckGo, one of Google’s search engine competitors, has also spoken out against FLoC, deactivating it on search pages and releasing a Chrome extension that prevents FLoC tracking across the web.

Why Has FLoC Failed?

All cookieless solutions are usually assessed based on:

  • their capacity to enable efficient marketing on the open web (the commercial angle),
  • their compliance with privacy rules (the legal angle),
  • and their impact on market openness and competitiveness (the competition angle).

Let’s have a look at where FLoC experienced issues.

FLoC Can’t Replace Third-Party Cookies

As per a claim by Google, FLoC could provide an effective replacement for cookie-based solutions, and advertisers could expect at least 95% of conversions on every dollar spent when compared to cookie-based advertising.

But when Criteo performed a test between the two, FLoC was nowhere near third-party cookies in terms of performance.


The reason is that FLoC can only work with Chrome. This means there’s no room for other browsers, devices, and offline data. Moreover, it has a very restricted focus and doesn’t support several common marketing tactics like ad sequencing and frequency capping.

FLoC Can’t Achieve Targeting

The FLoC cohorts developed by Google are pretty fixed. With these, it becomes impossible to measure the interest in a specific product. This, in turn, makes retargeting impossible. So although a broad interest in a topic can be captured, a search for a specific product cannot.

For example, “show this ad to people who searched for Chanel perfumes on Amazon at least once in a week” would turn into “show this ad to perfume lovers.”

In fact, cohorts based on a person’s previous website visits aren’t always indicative of future conversion behavior. And while a study by Google indicated that the FLoC API achieves higher precision than a random cohort assignment. That benchmark is useless because marketers don’t spend blindly or at random.

The Presence of FLoC Complicates Attribution

Since cohorts replace identity, attribution will become problematic in the Google environment.

So how will the marketers measure the effectiveness of campaigns?

Visibility metrics would outweigh the conversion metrics, decreasing the overall marketing effectiveness.

While other proposals of the Privacy Sandbox could fix this problem (for example, TURTLEDOVE could use device data for attribution measurement), it risks nullifying FLoC’s privacy benefits.

FLoC Doesn’t Protect Privacy

Here’s a fact – advertisers’ ability to identify customers across digital channels violates privacy laws.

While replacing IDs with anonymous cohorts can solve the problem, there’s still a catch. No companies will have the user data except for Google. Google would still have access to users’ history of cohorts and the raw data saved in the browser’s cache.

So FLoCs were still not more private than the cookie system. Cross-site tracking using FLoC IDs is still possible, for example, by observing browsing habits and cohort assignments over time. There’s nothing stopping advertisers from grouping people together based on their sexual orientation or ethnicity.

Topics API: Interest-Based Advertising (IBA)

Google announced a new proposal in its Privacy Sandbox initiative called Topics API after two years of negative criticism, claiming that it will facilitate interest-based advertising long after it sunsets third-party cookies in its Chrome browser in 2023.

What exactly is the Topics API?

In short, when someone browses the internet, Chrome keeps track of the types of websites they visit, such as news. Each website may have a maximum of three categories.

Chrome decides which five categories the user is most interested in each week and then adds a sixth at random to improve user privacy. Chrome keeps these six preferences for three weeks. There are 350 interest groups in total for the purposes of initial testing.

Will Advertisers Benefit from Topics?

Topics’ aggregate browsing labels will most likely not be relevant to brand advertisers. If Google is merely tracking people generally based on areas of interest, the reporting may be difficult to trust.

“Does Google really believe that in a week, the average person’s attention span is limited to only five topics?” “Most people will think about work, eating out, entertainment, commuting, possibly holidays, health and wellbeing in an average week”, said Farhad Divecha, MD of digital marketing agency AccuraCast.

Unless Google incorporates some method to overlook everyday themes and focus on new areas of interest, five topics will be too limiting. According to Divecha, this could be an issue for advertisers that wish to target consumers based on common topics.

FLEDGE API: A Remarketing Solution

And what happens if users choose not to use the Topics API? Will there be no interest or topic-based information for advertisers, and thus no reach or frequency data to rely on?

FLEDGE API, or First Locally-Executed Decision over Groups Experiment, another proposal from Privacy Sandbox, might hold some answers.

The functionality is intended to support remarketing by allowing the advertiser-defined interest groups connected with a browser to be controlled by the browser rather than the advertiser or ad tech platform. It’s possible that the frequency tracking mechanism will be the same. However, Google has not stated whether or not this is the case.

This is how FLEDGE works:

  1. A user visits an advertiser’s website.
  2. The user’s browser is prompted to add an interest group.
  3. The user goes to a website that sells advertising space.
  4. In the browser, an ad auction is held.
  5. From trusted servers, the vendor and participating purchasers receive real-time data.
  6. The winning advertisement is displayed.
  7. The outcome of the auction is announced.
  8. A click on an advertisement is recorded.

So What Should Advertisers and Marketers Do Now?

If you want to know whether or not you should make any immediate changes to your adtech strategy, the answer is no. For now, you can keep doing what you used to do. However, one thing is clear: third-party cookies don’t have a future.

The best idea right now is to try experimenting with the alternatives. Explore how you can modify your existing advertising strategy by involving the latest tech and marketing practices. One of such new additions to your list should be Topics API.

Also, you should keep a continuous watch on Topics as it’s still in the evolving stage. Reviews and feedback from stakeholders and experts are being used to improve and develop the current structure.

Originally published February 23, 2022 - Updated November 11, 2022
Mobile reading? Scan this QR code and take this blog with you, wherever you go.
Dionysia Kontotasiou
Dionysia Kontotasiou Convert's Head of Integration and Privacy, helping customers with technical queries.
Carmen Apostu
Carmen Apostu In her role as Head of Content at Convert, Carmen is dedicated to delivering top-notch content that people can’t help but read through. Connect with Carmen on LinkedIn for any inquiries or requests.

Start Your 15-Day Free Trial Right Now.
No Credit Card Required

You can always change your preferences later.
You're Almost Done.
I manage a marketing team
I manage a tech team
I research and/or hypothesize experiments
I code & QA experiments
Convert is committed to protecting your privacy.

Important. Please Read.

  • Check your inbox for the password to Convert’s trial account.
  • Log in using the link provided in that email.

This sign up flow is built for maximum security. You’re worth it!