Customer Cohort Analysis: How to Read Retention Curves and Act on Them
Quick answer
Customer cohort analysis groups customers by a shared start period (usually signup month) and tracks how each group retains over time, so you see whether retention is improving cohort-over-cohort and whether the retention curve flattens (a healthy sign of product-market fit) or keeps declining. It converts a single blended churn rate into a diagnostic: which cohorts leak, when they leak, and whether recent changes moved the curve.
A flat churn rate of 4% month-over-month sounds acceptable — until you realize your newest customers are churning at 10% while your tenured accounts are locked in. A blended rate hides two different realities. Cohort analysis pulls them apart.
Key takeaways
- Cohort analysis groups customers by start period (usually signup month) and tracks retention per group over time.
- A flattening retention curve signals product-market fit; a continuously decaying curve means leaks you need to plug.
- Logo cohorts (count of accounts) and revenue cohorts (dollar value) answer different questions — use both.
- Cohort analysis reveals whether product changes, onboarding tweaks, or pricing shifts actually improve retention, or just shift churn to later months.
What is customer cohort analysis?
Customer cohort analysis segments your customer base into groups defined by a shared start date — typically the month they first subscribed. Instead of a single churn rate, you get a time-series per cohort: how many from the January cohort were still active after 1 month, 3 months, 6 months, and so on.
The key insight is that you compare cohorts side by side. If the March cohort retains at 90% after month one but the July cohort retains at 85%, something changed — and you need to diagnose whether that change is within your control (onboarding, ICP shift) or external (seasonality, market pressure).
How to build a retention cohort? (step by step)
Building a retention cohort is straightforward, whether you use a spreadsheet, SQL, or a dedicated product analytics tool. Here are the steps:
1. Define the cohort dimension. Most often it's first subscription date, but you can also use first activated date, first payment, or account creation month. Stick to monthly cohorts for B2B SaaS because daily cohorts are noisy.
2. Define the retention event. What counts as "retained"? A customer who remains subscribed at the end of the period. For prepaid annual plans, you might use renewal as the event. For monthly billing, any month with an active subscription counts.
3. Create the cohort matrix. Each row is a cohort (e.g., January 2024), each column is a time period (Month 0, Month 1, Month 2…). Fill in the number of customers who are still active at each period.
4. Normalize to percentages. Divide each cell by the starting size of that cohort. This gives a retention curve from 100% down to your mature retention rate.
Example (three cohorts):
| Cohort | Month 0 | Month 1 | Month 2 | Month 3 |
|---|---|---|---|---|
| Jan 24 | 100 | 80 | 70 | 65 |
| Feb 24 | 110 | 85 | 72 | — |
| Mar 24 | 95 | 70 | — | — |
Convert the counts to retention percentages and the real story appears. At Month 1: Jan = 80%, Feb = 77% (85 of 110), Mar = 74% (70 of 95) — each new cohort retains slightly worse. At Month 2, February drops to 65% (72 of 110), below January's 70%. Newer cohorts are churning faster, not slower — exactly what a single blended churn number would hide.
How to read the retention curve (flattening vs decaying)
The shape of the retention curve is the most critical signal. Two patterns dominate:
Flattening curve: Retention drops sharply in the first few months (typical for any subscription) then stabilizes. A flat tail means customers who make it past the initial period stay indefinitely. This is a strong indicator of product-market fit — your product becomes stickier over time.
Decaying curve: Retention continues to decline period after period with no flattening. This suggests that every new cohort experiences the same erosion at each stage. The root cause may be a poor onboarding experience, mismatched target customer, or lack of ongoing value delivery.
To evaluate your own curves: overlay the last 4-6 cohorts on one chart. If the newer cohorts show a higher plateau than older cohorts at the same month, your changes are working. If curves are getting worse, you have a retention regression.
Logo vs revenue cohorts: which to use
Both matter, and they answer different questions. Here’s a comparison:
| Criterion | Logo Cohort | Revenue Cohort |
|---|---|---|
| What it tracks | Number of active accounts | Recurring revenue (usually MRR) |
| What it reveals | Product stickiness—are customers staying? | Monetary stickiness—are they expanding, downgrading, or fully leaving? |
| Sensitive to | Onboarding, engagement, churn triggers | Pricing, upselling, seasonality, expansion revenue |
| When to use | To diagnose early churn and product friction | To measure net revenue retention (NRR) and growth health |
For most B2B SaaS companies, you should track both. A company can have excellent account retention but poor revenue retention if customers are downgrading. Conversely, a few big expansions can mask account churn in revenue cohorts. Use logo cohorts for early-warning churn signals and revenue cohorts for financial forecasting.
What cohort analysis reveals that blended churn hides
Blended churn is a rearview mirror. Cohort analysis is a diagnostic x-ray. Here’s what you see with cohorts that you miss with a single rate:
- Early churn spikes. New customers often churn faster than old ones. Cohort analysis isolates that first-month drop-off. If your blended churn is 5% but new cohorts churn at 15% in month one, you have an onboarding problem.
- Impact of product or pricing changes. Launch a new feature or raise prices? Compare cohorts before and after the change. If retention improves or declines, you have a causal signal.
- Seasonal patterns. Cohort analysis reveals whether Q4 cohorts always churn more, allowing you to adjust sales compensation or support staffing.
- Improvement (or deterioration) over time. Are your recent cohorts retaining better than last year’s? If yes, your retention initiatives are working. If no, you need to revisit your approach.
Cohort analysis is not a one-time exercise. It should be a recurring report you review alongside your churn rate calculations and net revenue retention. Pairing it with industry benchmarks gives context for whether your curves are healthy or alarming.