Cohort Analysis

What is cohort analysis?

Cohort analysis (also called cohort tracking, cohort-based analysis, or vintage analysis) is a method of dividing customers into groups based on a shared attribute — most commonly the month or quarter they were acquired — and tracking how each group behaves over time. It answers questions that aggregate metrics cannot: is retention improving? Are newer customers spending more or less? Is churn getting better or worse with each new batch of customers?

Without cohort analysis, a company with 10,000 customers sees one churn rate number. That number blends the behavior of customers acquired three years ago (who are loyal and sticky) with customers acquired last month (who haven't yet decided whether to stay). The aggregate number masks the real trajectory. Cohort analysis strips away the blending and shows each group on its own timeline.

For B2B SaaS, the most common cohort structure is monthly acquisition cohorts tracked over 12-24 months. Healthy companies see retention curves that flatten — meaning each cohort stabilizes at a predictable retention rate after an initial drop-off period. A flattening curve at 75-85% after month 6 signals strong product-market fit. A curve that continues declining past month 12 signals a structural retention problem.

Cohort analysis differs from aggregate analysis in that it preserves time as a variable. Aggregate metrics report a single number for the entire customer base. Cohort analysis shows how that number evolved for each group — revealing whether the business is improving, degrading, or holding steady.

Why cohort analysis matters for operators

Operators who rely on aggregate retention numbers often discover problems too late. If last quarter's cohort is churning at 2x the rate of cohorts acquired 18 months ago, the aggregate number barely moves — because the older, larger cohorts dilute the signal. By the time the blended number reflects the problem, 3-4 months of bad cohorts have already been acquired.

Without cohort analysis, you see one retention number and assume the business is stable. With it, you see that the January cohort retained 78% at month 6, the March cohort retained 71%, and the May cohort is tracking toward 64%. The trend is visible months before it hits the aggregate.

A typical $3M ARR SaaS company tracking cohorts for the first time discovers that its most recent 3 months of customers retain at 8-12 percentage points below the 12-month average. The aggregate NRR still looks healthy because older cohorts are expanding. But the new cohort data reveals an onboarding or product-quality problem that, if left unaddressed, will depress NRR in 6-9 months.

How cohort analysis works

Cohort analysis follows a structured process rather than a single formula. Here is the standard approach with a retention curve example.

Step 1: Define the cohort

Group customers by acquisition month. January 2026 cohort = all customers who made their first purchase or started their subscription in January 2026.

Step 2: Choose the metric

Common metrics for cohort tracking: retention rate, revenue per customer, repurchase rate, cumulative spend, or LTV.

Step 3: Build the retention table

Example: Monthly Retention by Acquisition Cohort

Cohort     | M0    | M1   | M2   | M3   | M4   | M5   | M6
-----------+-------+------+------+------+------+------+------
Jan 2026   | 100%  | 82%  | 74%  | 70%  | 68%  | 66%  | 65%
Feb 2026   | 100%  | 79%  | 70%  | 66%  | 63%  | 61%  |  —
Mar 2026   | 100%  | 76%  | 67%  | 62%  | 59%  |  —   |  —
Apr 2026   | 100%  | 73%  | 64%  | 59%  |  —   |  —   |  —

Step 4: Read the curve

Each row is one cohort. Each column is time since acquisition. A healthy pattern shows retention dropping sharply in months 1-2, then flattening. In this example, January's curve is flattening around 65%. April's curve is steeper and lower — a signal that something changed in the product, onboarding, or customer quality.

Cohort retention benchmarks by company type

How retention curves compare across business models. Ranges based on industry data.

Sources: ChartMogul SaaS Retention Study 2025 (n=2,600), Recurly Subscription Benchmark Report 2025, industry-observed ranges.

Common mistakes when running cohort analysis

1. Using cohorts that are too large

A quarterly cohort blends January, February, and March customers. If a product change in February caused churn, the quarterly cohort dilutes the signal. Use monthly cohorts for operational decisions. Quarterly is acceptable for board-level reporting only.

2. Not controlling for cohort size

A cohort of 12 customers in a slow acquisition month produces noisy data. Small cohorts amplify individual behavior into apparent trends. Flag any cohort under 30 customers as statistically unreliable. Weight analysis toward cohorts with meaningful sample sizes.

3. Comparing cohorts at different maturity stages

The January cohort at month 6 is not comparable to the April cohort at month 2. Compare cohorts at the same point in their lifecycle. "January at M6 vs. December at M6" is valid. "January at M6 vs. April at M2" is not.

4. Ignoring revenue cohorts in favor of retention-only cohorts

A cohort that retains 70% of customers but those customers downgrade plans shows healthy retention and weak revenue. Track both logo retention (customer count) and revenue retention (NRR) by cohort. The gap between the two reveals expansion or contraction behavior.

5. Running cohort analysis once instead of continuously

A one-time cohort study shows a snapshot. The value compounds when you track every new cohort against the same framework. After 6-12 months of continuous tracking, you can predict a cohort's 12-month trajectory from its first 60 days of data.

How Fairview tracks cohort analysis automatically

Fairview's Margin Intelligence builds acquisition cohorts from your CRM (HubSpot, Salesforce, Pipedrive) and payment data (Stripe, Shopify) without manual spreadsheet work. Each cohort is tracked across retention, revenue, and margin — showing not just who stayed, but how much they contributed.

The Operating Dashboard displays a rolling cohort retention table updated automatically. You see month-over-month retention by cohort, revenue per cohort, and the flattening point where each group stabilizes. When a new cohort's retention trajectory deviates from the historical pattern, the Next-Best Action Engine flags it: "March 2026 cohort is retaining 9 points below January at the same lifecycle stage."

→ See how Margin Intelligence works

Cohort analysis vs aggregate analysis

People sometimes default to aggregate analysis because it is simpler. The trade-off is significant.

Aggregate analysis tells you where you are. Cohort analysis tells you where you are heading. Use aggregate metrics for weekly dashboards. Use cohort analysis for strategic decisions about retention, pricing, and customer quality.

FAQ

What is cohort analysis in simple terms?

Cohort analysis groups customers by when they signed up (or another shared trait) and tracks each group separately over time. Instead of one retention number for all customers, you see how each month's new customers behave as they age. It reveals whether newer customers are better, worse, or the same as older ones.

What is a good retention rate by cohort for B2B SaaS?

For mid-market B2B SaaS, a healthy cohort retains 78-88% of customers at month 6 and 70-85% at month 12. The retention curve should flatten after month 3-4. If the curve keeps declining past month 6, the product or onboarding process has a structural problem that aggregate numbers will not surface for months.

How do you run a cohort analysis?

Group customers by their acquisition month. Choose a metric — retention, revenue, or purchases. Build a table where each row is a cohort and each column is months since acquisition. Fill in the metric at each time point. Compare cohorts at the same lifecycle stage to identify whether things are improving or degrading.

What is the difference between cohort analysis and aggregate analysis?

Aggregate analysis reports one number for all customers. Cohort analysis separates customers into groups and tracks each over time. The key difference: aggregate metrics can look stable while new cohorts are deteriorating — because larger, older cohorts mask the trend. Cohort analysis reveals the direction of change before the aggregate catches up.

How often should you run cohort analysis?

Update cohort tables monthly. Review them in detail quarterly. For SaaS companies, monthly updates catch retention shifts within 60-90 days. For e-commerce, run repurchase cohort analysis monthly during promotional seasons and quarterly during steady-state periods.

What metrics should you track by cohort?

Five core metrics: logo retention rate, revenue retention (NRR), average revenue per customer, cumulative cohort LTV, and payback period. Tracking all five by cohort gives you a complete picture of customer quality — not just whether they stayed, but how much they contributed over time.

Related terms

• Cohort LTV — Lifetime value calculated per acquisition cohort rather than blended across all customers
• Churn Rate — Percentage of customers who cancel in a given period, the inverse of retention
• NRR (Net Revenue Retention) — Revenue retained from existing customers including expansion and contraction
• Customer Lifetime Value — Total revenue expected from a customer over the full relationship
• Repurchase Rate — Percentage of customers who make a second purchase, a key cohort metric for e-commerce

Fairview is an operating intelligence platform that tracks cohort analysis automatically alongside churn rate, NRR, and customer lifetime value. Start your free trial →

Siddharth Gangal is the founder of Fairview. He built automated cohort tracking into the platform after seeing operators make retention decisions on blended numbers that masked 3 months of deteriorating new-customer quality.