| name | cohort-analysis |
| description | Creates cohort analysis frameworks for understanding retention, revenue, and behavior patterns over time. Use when measuring how user groups perform across their lifecycle. |
| allowed-tools | Read Write Glob |
| metadata | {"author":"matthewhitcham","version":"1.0"} |
Cohort Analysis
When to Use This Skill
Use this skill when you need to:
- Understand customer retention trends by signup month
- Compare revenue performance across acquisition cohorts
- Identify which customer groups behave differently over time
- Build a repeatable cohort analysis framework for ongoing use
DO NOT use this skill for one-time snapshot metrics, individual customer analysis, or real-time monitoring. This is for longitudinal group-based analysis.
Core Principle
AVERAGES LIE — COHORT ANALYSIS REVEALS WHETHER YOUR BUSINESS IS ACTUALLY IMPROVING BY COMPARING HOW DIFFERENT GROUPS BEHAVE OVER THE SAME LIFECYCLE STAGE.
Phase 1: Brief
Required Inputs
| Input | What to Ask | Default |
|---|
| Analysis goal | "What are you trying to understand? (retention, revenue, engagement, churn)" | Customer retention |
| Cohort definition | "How should cohorts be grouped? (signup month, acquisition channel, plan tier)" | Signup month |
| Time granularity | "Weekly, monthly, or quarterly cohorts?" | Monthly |
| Observation window | "How many periods to track each cohort?" | 6 months |
| Data source | "Where is your customer data? (Stripe, CRM, database, spreadsheet)" | Spreadsheet or Stripe |
| Metric | "What metric per cohort? (active users, revenue, orders, logins)" | Active users (retention) |
GATE: Confirm brief before proceeding.
Phase 2: Design
Cohort Table Structure
Build a triangular matrix:
- Rows = cohorts (e.g., Jan signups, Feb signups)
- Columns = lifecycle periods (Month 0, Month 1, Month 2...)
- Cells = metric value or percentage
Analysis Framework
- Retention curve — how each cohort declines over time (or does not)
- Cohort comparison — are newer cohorts retaining better than older ones?
- Inflection points — where does the biggest drop-off happen?
- Stabilization — at what period do cohorts flatten out?
- Segment overlays — do specific segments (channel, plan, geography) retain differently?
GATE: Present the analysis framework and confirm before building templates.
Phase 3: Build
Deliverables
1. Cohort Analysis Template
- Spreadsheet or table structure ready for data entry
- Formulas for calculating retention percentages
- Conditional formatting rules (green = good retention, red = steep drop)
- Chart template for visualizing retention curves
2. Interpretation Guide
- How to read the cohort table (rows, columns, diagonal patterns)
- What "good" retention looks like for the business type
- Common patterns and what they mean:
- Improving rows = product is getting better
- Steep Month 1 drop = onboarding problem
- Flat curves after Month 3 = healthy core retention
3. Segmented Analysis Plan
- Which segments to overlay (acquisition source, pricing tier, geography)
- How to split cohorts for sub-analysis
- Comparison template for segment vs. segment
4. Action Recommendations Framework
- If Month 1 retention is below X%, focus on onboarding
- If retention flattens at Y%, invest in expansion over acquisition
- If newer cohorts are worse, investigate product or market changes
Phase 4: Polish
Reporting Cadence
- Monthly: update the cohort table with latest period data
- Quarterly: full analysis with segment overlays and trend narrative
- Trigger: run ad hoc analysis when a major change is made (pricing, onboarding, product)
Presentation Template
One-page summary with: cohort table, retention curve chart, top 3 insights, and recommended actions.
Example 1: SaaS Monthly Retention Cohorts
Cohorts by signup month. Metric: % of users active in each subsequent month. Goal: identify if onboarding improvements in March improved Month 1 retention vs. January cohort.
Example 2: E-commerce Revenue Cohorts
Cohorts by first purchase month. Metric: cumulative revenue per cohort. Goal: understand which acquisition channels produce higher lifetime value customers.
Anti-Patterns
- Using averages instead of cohorts — "Average retention is 40%" hides whether things are getting better or worse. Always break by cohort.
- Too many cohorts — 52 weekly cohorts on one chart is unreadable. Start monthly, go weekly only for specific investigations.
- Ignoring cohort size — a cohort of 5 users with 80% retention is not meaningful. Note sample sizes.
- Comparing different lifecycle stages — Month 3 of a January cohort and Month 1 of a March cohort are not comparable. Align by period.
- Analysis without action — cohort data is diagnostic. Always end with "So what do we do about it?"
Recovery
- Not enough data: Need at minimum 3 cohorts with 3 periods each. If data is thin, start tracking now and revisit in 3 months.
- No customer-level data: Use aggregate proxies (monthly active users vs. signups) for a rough cohort view. Plan to implement proper tracking.
- Retention looks terrible: Normalize expectations — B2C apps often see 20-30% Month 1 retention. Compare to industry benchmarks before panicking.
- User cannot interpret the table: Walk through one row in detail, explaining what each cell means for that specific cohort.
