FLoC and Its Deprecation
What FLoC Was Trying to Do
Federated Learning of Cohorts (FLoC) was Google's first major Privacy Sandbox proposal for interest-based advertising without third-party cookies, trialed in Chrome between 2021 and early 2022. Instead of tracking individuals, FLoC grouped users into cohorts of thousands of people with similar browsing patterns, computed entirely on-device using a clustering algorithm (SimHash) applied to browsing history, and exposed only a cohort ID to sites — not the browsing history itself. The idea was that advertisers could target a cohort ID the way they previously targeted an interest segment, while no individual's specific browsing history ever left their machine.
Why It Was Killed
FLoC collapsed under a combination of regulatory, technical, and competitive pressure. Privacy researchers demonstrated that cohort IDs, combined with other available signals (IP address, screen resolution, installed fonts), could still be used to re-identify or fingerprint individuals with meaningful accuracy — the clustering reduced but didn't eliminate the ability to narrow down a specific user. Regulators, especially the UK's CMA, scrutinized Google's dominant position in both the browser market and the ad market simultaneously defining the replacement standard, raising antitrust concerns about Google unilaterally deciding the rules of a market it also competes in. Browser competitors (Brave, Vivaldi, DuckDuckGo) blocked FLoC outright before Google itself abandoned it. Google formally discontinued FLoC in early 2022, replacing it with the Topics API as a redesigned, less granular successor.
FLoC vs Topics API
| Aspect | FLoC | Topics API (successor) |
|---|---|---|
| Grouping granularity | Thousands of fine-grained behavioral clusters | ~470 fixed, human-curated topic categories |
| Re-identification risk | Demonstrated to be fingerprintable in combination with other signals | Reduced via fixed taxonomy, random topic injection, capped observability |
| Transparency | Opaque cohort IDs with no public meaning | Published, human-readable taxonomy |
| Industry reception | Rejected by regulators and competing browsers | Cautious, partial adoption; urgency reduced after cookie deprecation reversal |
The Core Lesson: Privacy-Preserving ≠ Privacy-Proof
What MarTech Teams Actually Experienced
Very few advertisers built meaningful targeting infrastructure directly on FLoC cohort IDs before it was killed, since the trial was limited to a small percentage of Chrome traffic in a handful of regions and the specification was explicitly marked experimental throughout. The practical impact on most marketing stacks was minimal — the bigger lesson was reputational: it signaled early that any single-vendor-proposed replacement for third-party cookies would face intense scrutiny, and it set the expectation that identity and targeting strategy needed to diversify across multiple approaches (first-party data, clean rooms, contextual targeting) rather than betting on one browser vendor's cookie replacement.
// Historical reference only — FLoC is fully discontinued and this
// API no longer exists in any current Chrome version. Included here
// purely to illustrate what the deprecated interface looked like.
if (document.interestCohort) {
document.interestCohort().then(cohort => {
console.log(cohort.id, cohort.version);
// { id: "14159", version: "chrome.1.0" }
});
}Don't Build Durable Strategy on Any Single Browser Proposal
What's Next?
FLoC's replacement, the Topics API, inherited its goals but redesigned the mechanics around the specific criticisms that killed its predecessor, and understanding those design choices clarifies what's actually durable about the current approach versus what might still change.
I build these systems professionally.
Whether it's a RAG pipeline, analytics migration, or AI workflow — let's talk.