Cohort Analysis

Cohort Analysis

Perform cohort analysis on user engagement data — retention curves, feature adoption trends, and segment-level insights. Use when analyzing user

Category: development Source: phuryn/pm-skills

What Is This?

Overview

Cohort analysis is a behavioral analytics technique that groups users by a shared characteristic or event, then tracks how those groups behave over time. Rather than looking at aggregate metrics that blend all users together, cohort analysis isolates specific segments to reveal patterns that would otherwise remain hidden in the noise. This approach is foundational to understanding whether a product is genuinely improving user retention or simply masking churn with new user acquisition.

The technique works by defining a cohort, typically users who signed up, activated a feature, or completed a key action within the same time window, and then measuring their behavior across subsequent periods. The result is a structured view of how engagement evolves, decays, or stabilizes over days, weeks, or months. Retention curves, feature adoption timelines, and churn patterns all become measurable and comparable across different user segments.

This skill enables analysts and product managers to run cohort analysis on user engagement data systematically. It supports building retention curves, studying feature adoption trends, and generating segment-level insights that inform product decisions with precision.

Who Should Use This

  • Product managers who need to evaluate whether product changes are improving long-term user retention across different user segments
  • Data analysts responsible for building retention reports, interpreting behavioral trends, and presenting findings to stakeholders
  • Growth engineers investigating where users drop off in activation flows or which cohorts respond best to onboarding changes

Why Use It?

Problems It Solves

  • Aggregate retention metrics hide the difference between improving and declining cohorts, making it impossible to know if product changes are working
  • Feature adoption data without cohort segmentation cannot reveal whether new users adopt features differently than long-term users
  • Churn investigations without cohort context fail to identify when users leave, making root cause analysis guesswork
  • Blended averages obscure the performance of specific acquisition channels, making marketing attribution unreliable
  • Without cohort tracking, teams cannot measure the true impact of onboarding changes on downstream retention

Core Highlights

  • Generates retention curves segmented by cohort start date, acquisition channel, or user attribute
  • Tracks feature adoption rates across cohorts to identify adoption acceleration or stagnation
  • Supports weekly, monthly, and custom time-window cohort definitions
  • Produces heatmap-style retention tables for quick visual pattern recognition
  • Enables comparison of pre-change and post-change cohorts to measure product improvement
  • Identifies churn inflection points where engagement drops sharply within a cohort lifecycle
  • Supports segment-level filtering by plan type, geography, device, or custom user properties

How to Use It?

Basic Usage

A standard cohort retention query groups users by their signup month and measures the percentage still active in each subsequent month.

SELECT
  DATE_TRUNC('month', first_seen) AS cohort_month,
  DATE_TRUNC('month', event_date) AS activity_month,
  COUNT(DISTINCT user_id) AS active_users
FROM user_events
GROUP BY 1, 2
ORDER BY 1, 2;

This query forms the foundation of a retention table. Divide each month's active users by the cohort's initial size to get retention percentages.

Specific Scenarios

Scenario 1: Measuring onboarding changes. Compare the 30-day retention of cohorts who signed up before and after an onboarding redesign. If the post-change cohort retains at 45 percent versus 32 percent at day 30, the redesign produced a measurable improvement.

Scenario 2: Feature adoption by cohort. Track what percentage of each monthly cohort activates a specific feature within their first 14 days. Declining adoption rates in recent cohorts may indicate a discoverability problem introduced by a UI change.

Real-World Examples

A SaaS company notices overall retention looks stable at 60 percent month-over-month. Cohort analysis reveals that older cohorts retain at 75 percent while newer cohorts retain at only 45 percent, masked by the volume of new signups. This finding triggers an immediate onboarding audit.

A mobile app team uses cohort analysis to discover that users acquired through paid social churn 40 percent faster than organic users by day 7, prompting a reallocation of acquisition budget.

Important Notes

Requirements

  • Access to an event-level data store with user identifiers, event timestamps, and a defined first-seen or signup date per user
  • A consistent definition of "active" that is agreed upon before analysis begins, such as logging in, completing a core action, or generating revenue
  • Sufficient cohort sizes to produce statistically meaningful retention percentages, generally at least 100 users per cohort
  • A data warehouse or analytics tool capable of running grouped time-series queries, such as BigQuery, Redshift, Snowflake, or Mixpanel