Cohort-Based Advertising (Privacy Sandbox)
The Core Idea: Group, Don't Identify
Cohort-based advertising targets groups of users who share browsing interests rather than targeting an individual by a persistent ID. Google's Privacy Sandbox initiative built two relevant APIs on this principle: the Topics API (successor to the abandoned FLoC) and the Protected Audience API (formerly FLEDGE), which enables on-device remarketing-style auctions without a third-party ID leaving the browser.
The Topics API assigns each visited site to one or more of roughly 350+ predefined interest categories from a public taxonomy (e.g., 'Fitness' > 'Running & Jogging'), and the browser surfaces only a handful of a user's top topics from the last few weeks to any given caller — never the user's full history, and never a unique identifier.
How Topics API Differs from FLoC
FLoC clustered users into cohorts based on raw browsing history similarity using an algorithm (SimHash) that produced opaque numeric cohort IDs — critics quickly showed cohort IDs could act as a quasi-fingerprint and could leak sensitive inferences (e.g., a cohort correlating with a health condition or sexual orientation). Google deprecated FLoC in 2022 after regulatory and publisher pushback.
Topics API addresses this by using a fixed, human-curated, publicly auditable taxonomy instead of an opaque clustering algorithm, capping the categories to coarse, non-sensitive topics, rotating topics weekly, and limiting each caller's visibility to a small random sample of a user's top topics rather than a full profile.
Protected Audience Runs the Auction On-Device
Privacy Sandbox Cohort APIs at a Glance
| API | Replaces | What It Reveals to Callers |
|---|---|---|
| Topics API | Interest-based targeting cookies | A few coarse topic IDs from a fixed taxonomy, rotated weekly |
| Protected Audience API | Third-party cookie remarketing/retargeting | Only the final on-device auction outcome, not raw interest groups |
| Attribution Reporting API | Third-party cookie-based conversion tracking | Aggregated, noised event- and summary-level reports |
// Reading a user's browser-supplied topics (runs in a supporting browser context)
if ("browsingTopics" in document) {
const topics = await document.browsingTopics();
// topics: [{ topic: 243, taxonomyVersion: "2", modelVersion: "1", version: "chrome.1:2:1" }, ...]
topics.forEach((t) => {
sendToAdServer({ topicId: t.topic, taxonomyVersion: t.taxonomyVersion });
});
}Practical Adoption Status
Google ultimately decided to keep third-party cookies available in Chrome with expanded user choice controls rather than eliminating them outright, which took pressure off forced Privacy Sandbox adoption — but the APIs remain live, and Safari's and Firefox's independent cookie blocking means cohort-based approaches are still the only Google-native path to interest targeting on a meaningful share of non-Chrome traffic. Most ad tech vendors treat Privacy Sandbox integration as a hedge, not yet a primary targeting pillar, running it alongside first-party data and contextual strategies rather than instead of them.
Don't Over-Invest Based on Deadline Pressure Alone
What's Next?
Cohort-based advertising is one strand of the broader privacy-resilient targeting toolkit alongside contextual advertising and first-party identity — the strongest stacks blend several rather than betting on a single replacement for third-party cookies.
I build these systems professionally.
Whether it's a RAG pipeline, analytics migration, or AI workflow — let's talk.