Cohort Analysis

Behavioral analysis technique that groups users by shared characteristics or time periods, then compares outcomes across cohorts. Reveals how user segments respond differently to changes.

What is Cohort Analysis?

Cohort analysis groups users into cohorts based on shared characteristics (new vs. existing, premium vs. free, mobile vs. desktop) or time periods (users acquired in week 1, week 2, etc.), then compares how each cohort behaves over time. Rather than averaging behavior across all users (which masks variation), cohort analysis reveals whether different user segments respond differently to the product, features, or changes. A change might improve average metrics while harming a specific segment—only cohort analysis reveals this.

The power is segmentation-level insight. An A/B test shows whether a change improved conversion (5% overall), but cohort analysis shows which segments benefited (paid users: +10%, free users: -2%) and which were harmed. This reveals whether the change is universally good or good for some segments while degrading experience for others.

Time-Based vs. Behavioral Cohorts

Time-based cohorts group users by acquisition date: “users who signed up in March 2024” vs. “users who signed up in April 2024.” This reveals cohort effects—do different acquisition cohorts behave differently? New cohorts might have higher churn (seasonal effects, different marketing messages), or they might behave similarly (suggesting consistent product experience). Tracking cohort behavior over time creates a retention curve—how long do users stay engaged after acquisition?

Behavioral cohorts group users by actions: “users who completed onboarding” vs. “users who never onboarded,” “users who have 10+ analyses” vs. “users with fewer than 10.” This reveals correlation between behavior and retention or engagement. Does completing onboarding predict higher retention? Does reaching certain usage milestones predict lower churn?

Cohort Tables and Retention Curves

Cohort tables are the standard visualization: rows are acquisition cohorts, columns are weeks or months, and cells show retention percentage. This reveals patterns: Do early cohorts retain better than recent cohorts? Is there a cliff at month 3? These patterns often point to explanations: a product launch changed onboarding quality, a pricing change reduced value perception, a bug harmed a specific cohort.

Retention curves—plotting cohort retention over time—make trends visible. A cohort with steep dropoff (50% churn in first week) suggests onboarding problems or product-market misalignment. A cohort with gentle decline (stable 10% monthly churn) suggests satisfied users with natural attrition.

Segmentation: Which Cohorts Matter?

Cohort analysis creates analysis debt if you examine every possible segmentation. The discipline is identifying which cohorts matter for your business: if you have a two-tier pricing model, segment by tier. If geography affects experience, segment by region. If customer size drives behavior, segment by company size. Define meaningful cohorts, not arbitrary ones.

Why It Matters for Product People

Cohort analysis prevents the averaging fallacy: assuming population-level metrics hide important segment-level variation. A retention improvement of 5% is great if it affects all users equally; it’s a warning sign if it’s driven by one segment while others regress.

For executives, cohort analysis informs resource allocation. If mobile users have lower retention, mobile improvements become a priority. If a specific customer size has high churn, adjusting positioning or product focus for that segment becomes strategic.

Cohort analysis complements A/B testing (which measures population-level effect) by revealing segment-level variation. It’s foundational to retention analysis and funnel analysis. Regular cohort reviews (monthly or quarterly) track how product changes affect different user segments.