Why Cohort Analysis Is the Key to Predicting Churn
Most SaaS founders look at a single churn rate and assume they understand their retention story. They don't. A blended monthly churn rate of 3% could be hiding the fact that customers acquired through paid ads churn at 7% while referral customers churn at just 1.5% — a difference that completely changes your unit economics and where you should invest your growth budget.
Cohort analysis separates customers into groups — typically by acquisition month — and tracks how each group behaves over time. When done correctly, it transforms churn from a lagging indicator into a leading one. You can see churn coming months before it hits your MRR, identify which customer segments are accelerating toward cancellation, and intervene while there's still time.
This guide covers how to build a cohort analysis from scratch, the formulas that matter, how to read the signals that predict churn before it happens, and what benchmark retention curves look like at each stage of SaaS growth.
The Cohort Retention Grid: Your Core Analytical Tool
The cohort retention grid is the foundation of all cohort analysis. It answers one question for every group of customers: of those who started in month X, what percentage are still paying in month X+N?
Building a Cohort Retention Grid
To construct a cohort retention grid, you need three data points for each customer: their first payment date, a monthly activity record, and their churn date (if applicable).
Group customers by signup month. For each cohort, calculate retention at each time horizon:
Retention Rate (Month N) = Customers still active at Month N / Customers who signed up in Cohort Month × 100
A simple grid looks like this:
| Cohort | Month 0 | Month 1 | Month 2 | Month 3 | Month 6 | Month 12 |
|---|---|---|---|---|---|---|
| Jan 2026 | 100% | 94% | 88% | 84% | 76% | 69% |
| Feb 2026 | 100% | 92% | 86% | 81% | 73% | ? |
| Mar 2026 | 100% | 91% | 84% | 79% | ? | ? |
| Apr 2026 | 100% | 93% | 87% | ? | ? | ? |
Each row is a cohort. Each column is months since signup. Reading down any column tells you whether newer cohorts are retaining better or worse than older ones — this is how you measure whether your product and onboarding improvements are actually working.
Revenue Cohort Grid vs. Customer Cohort Grid
Most teams start with customer cohort grids (tracking customer count retention), but revenue cohort grids are often more informative. A revenue cohort grid tracks what percentage of the original MRR from each cohort is still active — and if customers expand, the percentage can exceed 100%.
Revenue Retention Rate (Month N) = MRR from cohort still active at Month N / Starting MRR from cohort × 100
If January's 50 customers started at $200 MRR each ($10,000 total) and by Month 12 you still have $8,500 from that cohort — some churned, some expanded — your revenue retention is 85%. A cohort with 72% customer retention but 85% revenue retention tells you your retained customers are expanding, which is a healthy sign.
The Cohort Retention Curve: Reading the Shape
Every SaaS cohort has a retention curve — the shape of the line as you plot retention over time. Learning to read that shape is more valuable than any single retention percentage.
The Healthy Retention Curve
A healthy SaaS retention curve looks like a ski slope that levels off. It drops steeply in the first 1-3 months (initial churn from customers who don't reach value), then flattens and stabilizes. The key diagnostic question: does the curve eventually reach a plateau?
If your 12-month-old cohort has 65% retention and your 24-month-old cohort also has 65% retention, that's a sign the remaining customers are committed — they've found value and aren't going anywhere. That plateau is your loyal core, and its size determines your long-term LTV.
The Declining Curve (No Plateau)
If your cohort retention keeps declining without flattening — Month 12 at 50%, Month 18 at 40%, Month 24 at 30% — you have a structural retention problem. There is no loyal core. Every customer has roughly equal probability of churning at any given time, which means there's no natural floor on churn. This pattern is common in products that solve a one-time problem rather than creating ongoing habit.
The fix is usually product depth: adding features that create ongoing dependency, usage patterns that compound in value, or integrations that make switching painful.
The Expanding Curve (NRR > 100%)
The best SaaS businesses have cohort revenue curves that eventually trend *above* 100%. This happens when expansion revenue from surviving customers outpaces churn losses from departing ones. If Month 12 revenue retention is 108%, your retained customers are growing their spend faster than you're losing customers.
This is the pattern that drives Net Revenue Retention (NRR) above 100%. For more on NRR benchmarks and what drives them, see our full NRR guide.
How Cohort Analysis Predicts Churn Before It Happens
The most powerful application of cohort analysis isn't looking backward — it's using patterns in early cohort behavior to predict where a cohort is headed before the churn actually occurs.
Month 1 Retention as a Leading Indicator
Research across thousands of SaaS companies shows that Month 1 retention is one of the strongest predictors of 12-month retention. If a cohort retains 95% through their first month, they'll typically retain 70-75% at Month 12. If they retain only 85% through Month 1, Month 12 retention typically falls to 55-60%.
This means you don't need to wait a year to know if a cohort is on a healthy trajectory. A poor Month 1 retention rate for your March cohort tells you, in April, that those customers are at elevated churn risk — and you have 10 more months to intervene.
Month 1 Retention Benchmarks:
Behavioral Signals Within a Cohort
Not all customers in a cohort are equal. Within any cohort, some customers will be on a path to long-term retention and others will churn within 90 days. The difference shows up in behavior before it shows up in cancellation data:
Activation rate. Customers who complete key setup steps within their first 7 days (connect integrations, invite teammates, upload data) retain at dramatically higher rates. Define your activation event and track it per cohort.
Feature adoption depth. Customers using 3+ core features in Month 1 churn at roughly half the rate of customers using only one feature. When a new cohort shows low feature adoption breadth, you can predict elevated 90-day churn.
Session frequency. Weekly logins versus monthly logins in the first 30 days predict 6-month retention with high accuracy. A cohort with low early session frequency is signaling that the product hasn't become habitual.
Support ticket sentiment. A cohort generating high frustration-coded support tickets in Month 1 (setup problems, feature gaps, billing confusion) has measurably higher churn through Month 3. Track ticket sentiment per cohort alongside retention metrics.
The Cohort Divergence Signal
One of the most actionable signals in cohort analysis is cohort divergence: when a new cohort's early retention is tracking meaningfully below previous cohorts at the same time horizon.
Example: Your January and February cohorts both retained 93% through Month 1. Your March cohort is only retaining 86% through Month 1. Something changed. It could be a product issue introduced in March, a change in your acquisition channel bringing in less-fit customers, a pricing change, or a competitor action.
Detecting this divergence a month after the March cohort signed up gives you weeks or months to investigate and fix the problem before it compounds into a serious MRR impact. Without cohort analysis, you'd see the blended churn rate tick up slightly in June — three months later, with no clear cause.
Churn Prediction Formula: Estimating Future MRR Loss
Once you have cohort retention curves, you can build a bottom-up churn prediction model. Here's the formula:
Predicted Churned MRR (Month N) = Σ (Cohort MRR × Expected Cohort Churn Rate at Month N)
Where expected cohort churn rate at Month N is derived from your historical retention curves for cohorts at the same age.
Step-by-Step Churn Forecast Example
Suppose you have three active cohorts:
| Cohort | Current MRR | Cohort Age | Observed Churn/Month at This Age |
|---|---|---|---|
| Jan | $32,000 | 6 months | 1.8% (based on Jan cohort history) |
| Mar | $28,000 | 4 months | 2.4% (higher early churn is typical) |
| May | $21,000 | 2 months | 3.5% (new cohort still in steep drop) |
Predicted next-month churn:
This is more accurate than applying a single blended churn rate because it accounts for the fact that newer cohorts churn faster than older, more stable cohorts. A blended 2.5% rate applied to $81,000 total MRR would predict $2,025 — close in this case, but the error compounds over multi-month forecasts and grows as your cohort age mix shifts.
For a full walkthrough of how this feeds into MRR forecasting, see the MRR forecasting model guide.
Segmenting Cohorts for Deeper Churn Insight
Simple time-based cohorts are the starting point. Once you're comfortable with the basics, segment your cohorts to identify which customer types drive healthy retention and which drive churn.
Acquisition Channel Cohorts
Group cohorts by how customers found you: organic search, paid ads, referrals, direct sales, partnerships. Channel cohorts almost always reveal dramatic retention differences. Referral and word-of-mouth cohorts typically retain 15-25 percentage points better at Month 12 than paid acquisition cohorts — because customers who came through a trusted recommendation arrived with better expectations and higher intent.
Plan Tier Cohorts
Customers on your highest-priced plan churn at lower rates in most SaaS businesses. This is partially selection (higher-ACV customers have more at stake) and partially product depth (they're using more features, creating more switching costs). If your Enterprise tier retains 85% at Month 12 while your Starter tier retains 52%, that's a strategic signal about where to invest in retention programs.
Company Size Cohorts
SMB customers churn at 3-5x the rate of enterprise customers in most B2B SaaS products. If you serve both segments, keeping them in separate cohorts prevents the low-churn enterprise segment from masking the high-churn SMB segment in your aggregated numbers.
First-Use Pattern Cohorts
Behavioral segmentation is the most predictive. Customers who completed your full onboarding flow in their first week are typically a different retention cohort from customers who skipped it — even when acquired from the same channel and on the same plan. Build these behavioral cohorts and track them separately to measure onboarding ROI directly.
Cohort Analysis Benchmarks: What Good Looks Like
Benchmarks vary significantly by ACV (average contract value) and customer type, but these ranges reflect common patterns across B2B SaaS:
Customer Retention by Stage and Segment
| Segment | Month 3 | Month 6 | Month 12 |
|---|---|---|---|
| SMB (<$100/mo ACV) | 75–82% | 65–72% | 55–65% |
| Mid-Market ($100–$1K/mo) | 83–89% | 75–82% | 68–76% |
| Enterprise (>$1K/mo) | 91–95% | 86–92% | 80–88% |
Revenue Retention Benchmarks
Revenue retention runs 5-10 percentage points higher than customer retention in healthy SaaS businesses, because churning customers tend to be smaller accounts while retained customers tend to expand.
| Stage | Month 6 Revenue Retention | Month 12 Revenue Retention |
|---|---|---|
| Early (< $2M ARR) | 72–80% | 65–75% |
| Growth ($2M–$20M ARR) | 80–88% | 75–82% |
| Scale (> $20M ARR) | 86–94% | 82–92% |
Top-decile SaaS businesses at scale show month-12 revenue retention above 100%, meaning expansion more than offsets churn. Companies like Snowflake and Datadog run month-12 revenue cohort retention above 130% — every cohort is worth more a year later than it was at inception.
Using Your MRR Calculator to Model Cohort Impact
Cohort analysis tells you what's happening. Your MRR model tells you what it means for your business. The connection is direct: improving your Month 12 cohort retention by 5 percentage points — from 65% to 70% — has a compounding revenue impact that typically represents 3-5x the revenue impact of acquiring 5% more new customers.
To see this in action for your own numbers, use the MRR calculator to model the revenue difference between your current retention curve and a 5-point improvement scenario. The results consistently surprise founders who've been focused primarily on acquisition.
Building a Cohort Analysis System: Practical Steps
Step 1: Extract the Data
You need a customer-level dataset with signup date, monthly payment amounts, and churn date. Export this from your billing system (Stripe, Chargebee, Recurly) or your CRM. Most billing platforms have cohort reports built in — check there before building custom queries.
Step 2: Build the Grid in a Spreadsheet
For teams under 1,000 customers, a spreadsheet cohort grid is entirely sufficient. Create one row per cohort month, one column per month-since-signup, and populate with retention percentages. Update it monthly — this 30-minute monthly task is one of the highest-ROI analytical activities in SaaS.
Step 3: Define Your Activation Event
Within-cohort behavioral analysis requires defining your activation event — the specific in-product action that predicts long-term retention. Common activation events: connecting an integration, inviting a second user, completing a core workflow. Identify yours by comparing what retained customers did in Month 1 that churned customers didn't.
Step 4: Set Up Cohort Alerts
Create an alert that fires when a new cohort's Month 1 retention falls more than 3 percentage points below your trailing average. This is your early warning system for product or acquisition problems.
Step 5: Review Cohorts in Your Monthly Metrics Meeting
Don't let cohort data become a monthly report nobody reads. Make it a standing agenda item: which cohorts are tracking above or below historical patterns, and what's the team doing about it?
Conclusion: Cohort Analysis as a Churn Early-Warning System
Blended churn rates tell you what happened. Cohort analysis tells you why, who, and — most importantly — what's about to happen.
The companies that build early churn prediction capabilities don't just react to churn faster than their competitors — they prevent significant portions of it entirely. A customer who shows behavioral signals of disengagement in Month 2 can still be saved by a well-timed success call, a personalized training session, or a product fix. A customer who submits a cancellation request in Month 8 is gone.
Start with a monthly cohort retention grid. Add revenue retention alongside customer retention. Track Month 1 retention as your leading indicator. Segment by acquisition channel and plan tier once your volume supports it. Review cohort trends monthly.
The data required to do this lives in your billing system right now. The only thing between you and a predictive churn model is the decision to build one.