What Is Cohort-Based Advertising?
Cohort-based advertising is a targeting method that groups users by a shared behavior, attribute, or timeline and serves ads to the entire group rather than to individually identified profiles. Instead of tracking a single user across the web, the approach assigns that user to a cohort, such as “users who visited a car comparison page in the last 14 days.” That cohort is then targeted as a unit, preserving advertising effectiveness while reducing reliance on individual-level identifiers.
The model gained significant traction after Google announced the deprecation of third-party cookies and proposed its Privacy Sandbox initiative, which includes the Topics API. Under that system, a browser assigns users to interest-based cohorts locally, on-device, and shares only the cohort label with advertisers, not the identity behind it.
How Cohort-Based Advertising Works
The mechanics follow four steps:
- Signal collection. The system collects behavioral signals: page visits, purchase history, video completions, and search queries.
- Cohort assignment. An algorithm groups users by shared signals within a defined lookback window, typically 7 to 30 days.
- Ad delivery. The DSP (demand-side platform) bids on impressions for users carrying a specific cohort label rather than a user ID.
- Performance aggregation. Platforms report results at the cohort level, not the individual level, limiting re-identification risk.
The lookback window is a critical lever. A 7-day cohort of users who searched for “running shoes” captures high purchase intent. A 30-day cohort of the same query captures a broader, cooler audience. Shorter windows generally produce higher CPAs but stronger conversion rates.
Cohort Types
Behavioral Cohorts
Formed around actions: product page views, cart abandonment, video watch percentage, or search query patterns. A retailer running a behavioral cohort of “cart abandoners in the last 10 days” is performing a form of retargeting without relying on a persistent cookie.
Temporal Cohorts
Grouped by acquisition date. A SaaS brand might target users who signed up for a free trial in the same calendar week, then serve them upgrade ads timed to the end of the trial period. This is common in lifecycle marketing and pairs naturally with audience segmentation.
Contextual Cohorts
Built from content consumption patterns rather than declared identity. A user who regularly reads articles about electric vehicles gets assigned to an auto-interest cohort. The Topics API uses this model: the browser classifies browsing history into one of several hundred interest categories and reports only the top topics to ad buyers.
Predictive Cohorts
Generated by machine learning models that score users by likelihood to convert, churn, or upgrade. Google’s Audience Expansion feature and Meta’s Advantage+ audiences both apply predictive cohort logic to extend reach beyond seed lists while maintaining targeting relevance.
Cohort Size and the Anonymity Threshold
Cohort-based systems require a minimum group size to prevent re-identification. Google’s original FLoC (Federated Learning of Cohorts) proposal required cohorts of at least 2,000 users before advertisers could see a cohort ID. The Topics API does not expose cohort IDs directly but operates on a similar principle: the system reports topics only if enough users share them.
Advertisers can estimate the effective reach of a cohort using this formula:
Cohort Reach = (Total Addressable Audience) × (Cohort Qualification Rate)
For example, if a brand’s addressable audience is 4 million users and 6% have visited a product page in the last 14 days, the cohort reach is approximately 240,000 users. A cohort that small may still qualify for targeting but will exhaust frequency caps quickly, requiring refresh cycles or cohort expansion.
Cohort-Based Advertising vs. Individual Targeting
| Dimension | Individual Targeting | Cohort-Based Targeting |
|---|---|---|
| Identifier required | Cookie, IDFA, email hash | Cohort label or topic category |
| Privacy exposure | High (cross-site tracking) | Lower (group-level signal) |
| Personalization depth | High | Moderate |
| Scale | Limited by ID match rates | Broader, no match rate floor |
| Reporting granularity | User-level | Aggregate only |
The trade-off is real: cohort-based approaches sacrifice granularity in exchange for scale and regulatory resilience. Brands running campaigns under GDPR or CCPA often find cohort targeting easier to justify from a consent standpoint than persistent cross-site tracking. This is part of why first-party data strategies have become central to building cohort seed audiences inside walled gardens like Google and Meta.
Real-World Applications
Retail and E-Commerce
Walmart Connect and Amazon DSP both offer cohort-based targeting rooted in purchase behavior. Amazon segments shoppers into categories such as “purchased pet supplies in the last 30 days” and surfaces those cohorts to pet food brands for upper-funnel display buys. Amazon DSP reported in 2023 that purchase-based audience segments drove a median 29% higher ROAS compared to contextual targeting alone on its network.
Streaming and Entertainment
Netflix, through its ad-supported tier launched in late 2022, sells cohort-based inventory grouped by genre consumption: thriller watchers, documentary viewers, and similar clusters. Advertisers cannot target individual subscribers but can reach genre cohorts with a minimum size threshold, which Netflix has not publicly disclosed.
B2B Advertising
LinkedIn’s Matched Audiences product allows advertisers to build cohorts from job title clusters, company size ranges, and skill endorsements. A cybersecurity vendor might define a cohort as “IT decision-makers at companies with 500 to 5,000 employees who engaged with cloud security content in the last 30 days.” This blends temporal and behavioral signals without accessing individual browsing history outside LinkedIn.
Measuring Cohort Campaign Performance
Because cohort-based campaigns report at the group level, attribution requires adjusted measurement frameworks. Incrementality testing and media mix modeling become more relevant than last-click attribution. A common approach is cohort holdout testing:
Incremental Lift = (Conversion Rate, Exposed Cohort) – (Conversion Rate, Holdout Cohort)
If an exposed cohort converts at 4.2% and a matched holdout cohort converts at 2.8%, the incremental lift is 1.4 percentage points, or roughly 50% relative lift. This method requires statistical significance, typically demanding cohort sizes above 10,000 users per arm to detect moderate effect sizes reliably.
Cohort-based advertising fits within the broader toolkit of programmatic advertising and complements behavioral targeting strategies for brands building privacy-resilient audience frameworks.
Key Takeaways
- Cohort-based advertising targets groups sharing a behavior or attribute, not individual user profiles.
- Cohort size thresholds (often 2,000 or more users) protect against re-identification and are required by systems like Google’s Privacy Sandbox.
- Lookback windows directly affect intent quality: shorter windows yield higher intent, smaller scale.
- Performance measurement shifts from user-level attribution to incrementality testing and aggregate reporting.
- Major platforms including Amazon DSP, LinkedIn, and Netflix’s ad tier already operate cohort-based models at scale.
Frequently Asked Questions About Cohort-Based Advertising
What is cohort-based advertising?
Cohort-based advertising is a digital ad targeting method that groups users by shared behaviors, attributes, or time periods and serves ads to the entire group rather than to individual profiles. It reduces reliance on cookies and personal identifiers while maintaining audience targeting at scale.
How is cohort-based advertising different from behavioral targeting?
Behavioral targeting traditionally relies on individual user-level identifiers such as cookies or device IDs to track and target specific people. Cohort-based advertising replaces the individual identifier with a group label, so advertisers see only aggregate behavior patterns, not individual browsing histories. The practical targeting effect is similar, but the privacy exposure is lower.
What is Google’s Topics API and how does it relate to cohorts?
The Topics API is Google’s Privacy Sandbox mechanism for interest-based advertising without cookies. A user’s browser classifies their browsing history into one of several hundred interest topics and shares only those topic labels with ad buyers, not the user’s identity or full history. It is a cohort-based approach where the assignment happens on-device.
What cohort size is needed for advertising campaigns?
Most cohort-based systems require a minimum group size to prevent re-identification. Google’s original FLoC proposal required at least 2,000 users per cohort before surfacing a cohort ID. For incrementality testing, cohort sizes above 10,000 users per arm are typically needed to detect moderate effect sizes with statistical confidence.
Can cohort-based advertising work without third-party cookies?
Yes. Cohort-based advertising is specifically designed to function without third-party cookies. Platforms like Amazon DSP, LinkedIn, and Netflix’s ad tier already operate cohort models using first-party behavioral data and contextual signals rather than cross-site tracking cookies.