Your overall retention rate is lying to you. Not maliciously, but structurally. It blends customers who signed up three years ago with customers who signed up last Tuesday, then hands you one number that describes neither group.

Retention cohort analysis fixes that. It groups customers by when they started, then tracks each group separately over time. Instead of one blended number, you see exactly how the customers you acquired in January behaved in month one, month two, and month six, right next to the customers you acquired in February. The pattern that emerges tells you where churn actually begins, whether your product changes are working, and which acquisition channels bring customers who stay.

This guide covers how to build a retention cohort table from scratch, how to read it the way an analyst does, what good retention curves look like for SaaS, and the part most articles skip: what to actually do once the table shows you a problem.

What Is Retention Cohort Analysis

A cohort is a group of customers who share a starting point. In retention analysis, that starting point is almost always the signup month or week. Everyone who signed up in March 2026 is the March cohort. Everyone who signed up in April is the April cohort.

Retention cohort analysis tracks what percentage of each cohort is still active in each period after that starting point. Month 0 is the signup month, so retention there is 100 percent by definition. Month 1 shows how many came back. Month 3 shows how many survived the early drop. Month 12 shows who truly stuck.

The output is usually a triangular table. Rows are cohorts, columns are periods since signup, and each cell holds a retention percentage. Newer cohorts have fewer filled columns because less time has passed for them, which is why the table slopes down to the right like a staircase.

Why bother with the extra structure? Because a blended retention number hides cause and effect. Suppose your overall monthly retention improved from 88 to 91 percent. Did your product get stickier, or did you just acquire a big batch of new customers who have not had time to churn yet? A blended number cannot answer that. A cohort table can, in about ten seconds.

The same logic powers customer churn analysis generally, but cohorts add the time dimension that makes findings actionable. You stop asking “what is our churn rate” and start asking “in which month of the customer lifecycle does churn happen, and is that changing.”

Cohort Analysis vs. Retention Rate: Why the Blended Number Misleads

It helps to be precise about what each number can and cannot tell you.

Your churn rate or its mirror image, retention rate, is a snapshot. It says: of the customers we had at the start of the period, this share remained at the end. It is useful for board decks and quick health checks, and we cover the relationship between the two in churn rate vs. retention rate.

The snapshot fails in three common situations.

Fast growth masks bad retention. New customers churn at much higher rates than tenured ones. If you are growing quickly, your customer base is dominated by new accounts, so blended churn looks high even if long-term retention is excellent. The reverse is worse: a slowing company can show improving blended retention purely because its base is aging, while every new cohort it acquires retains worse than the last.

Changes take months to show up. Ship a better onboarding flow today and blended retention barely moves, because the blended number is dominated by existing customers who never see the new flow. In a cohort table, the very next cohort shows you whether the change worked.

You cannot locate the problem in time. Blended churn of 5 percent per month tells you nothing about when in the lifecycle customers leave. Cohort analysis shows you whether you lose them in month 1 (an onboarding problem), around month 12 (a renewal problem), or steadily forever (a value problem). Each of those needs a completely different fix.

How to Build a Retention Cohort Table Step by Step

You can build a workable cohort table in a spreadsheet with data you already have. Here is the process.

Step 1: Define What “Retained” Means

This decision shapes everything downstream, and teams get it wrong more often than any other step.

For a subscription business, the honest definition is usually still paying: the account has an active subscription in that period. Activity-based definitions (logged in at least once) flatter your numbers, because plenty of customers log in during the month they cancel.

Decide two things explicitly:

  • The unit. Accounts for B2B, users for B2C. Mixing them corrupts the table.
  • The event. Payment retention for revenue questions, activity retention for product engagement questions. Run both if you can; the gap between them is itself informative, because an account that stops logging in but keeps paying is churn that has not happened yet.

Step 2: Pick the Cohort Grain

Monthly cohorts are the default for SaaS with monthly billing. Weekly cohorts make sense for products with fast onboarding loops or for reading the effect of a specific launch. Quarterly cohorts suit enterprise products that close a handful of deals per month, where monthly cohorts would be too small to read.

A practical floor: each cohort should contain at least 30 to 50 customers. Below that, one or two cancellations swing the percentage so hard that the table turns to noise. If your monthly cohorts are smaller than that, switch to quarterly.

Step 3: Assemble the Data

You need three fields per customer: an identifier, a start date, and either a churn date or a way to derive active status per period. Every billing system exports this. From there:

  1. Assign each customer to a cohort by start month.
  2. For each cohort, count how many were active in month 0, month 1, month 2, and so on.
  3. Divide each count by the cohort’s starting size.

Step 4: Lay Out the Triangle

Rows are cohorts, columns are months since signup. A simplified example:

CohortSizeM0M1M2M3M6
Jan 2026210100%68%61%57%51%
Feb 2026245100%71%64%60%54%
Mar 2026260100%74%68%63%n/a
Apr 2026258100%79%73%n/an/a
May 2026291100%80%n/an/an/a

Add a heatmap (green high, red low) and the table becomes readable at a glance. Empty cells simply mean not enough time has passed.

Step 5: Add a Revenue View

Logo retention counts accounts. Net revenue retention weighs them by what they pay, including expansion and downgrades. Run the same triangle on MRR and you will often find the two tables disagree: losing many small accounts while your large accounts expand produces ugly logo retention and healthy revenue retention. That is a very different business problem than the reverse. Our guide to net revenue retention covers the calculation in detail.

How to Read a Cohort Table Like an Analyst

The table has three reading directions, and each answers a different question.

Read Down the Columns: Are We Getting Better?

Fix a column, say M1, and read down. Each row is a newer cohort. If M1 retention climbs as you go down (68, 71, 74, 79, 80 in the example above), newer customers are surviving the first month better than older ones did. Something you changed is working: onboarding, targeting, pricing, activation.

If the column deteriorates as you read down, sound the alarm early. Worsening new-cohort retention is invisible in blended numbers for months, because new cohorts are a small share of the total base. The cohort table is often the only place this shows up before it hits revenue.

One caution: check cohort composition before celebrating or panicking. A cohort acquired during a heavy discount campaign, a viral spike, or a seasonal burst behaves differently for reasons that have nothing to do with your product. Annotate the table with what marketing was doing that month.

Read Across the Rows: When Do We Lose People?

Follow a single cohort left to right and you get its survival curve. The shape matters more than any single value.

  • A cliff between M0 and M1 means customers sign up and fail to reach value. This is an onboarding and activation problem. Look at your customer onboarding flow and your time to first value.
  • A steady downward slope that never flattens means customers reach value but the value does not compound. The product is useful once, not habitually. This is a product adoption problem: users never take up the features that make the product part of their weekly routine.
  • A curve that flattens after a few months is what good looks like. The customers who remain have integrated the product into how they work, and each additional month costs them little to renew.
  • A sharp step at month 12 in an annual-contract business is renewal churn. Customers who quietly disengaged months earlier all exit at the contract boundary. The cohort table shows the step; your usage data shows the disengagement that preceded it.

Read the Diagonals: What Happened That Month?

Cells on the same diagonal share a calendar month. If every cohort dips in the same calendar period regardless of age, the cause is external or global: a price increase, an outage, a botched release, a payment provider migration, seasonality. A diagonal pattern absolves the lifecycle and points at the calendar.

This is also where involuntary churn shows up. A wave of failed payments (an expired-card batch, a gateway change) hits all cohorts simultaneously, painting a diagonal stripe. If you see one, check payment failure rates before blaming the product; the fix is usually dunning management, not a roadmap change.

What Good Looks Like: SaaS Retention Benchmarks for Cohorts

Benchmarks vary by market, price point, and customer size, so treat these as orientation rather than targets. General figures for B2B SaaS, consistent with the ranges we compiled in our B2B SaaS churn rate benchmarks:

  • M1 retention. SMB-focused products commonly keep 60 to 80 percent of a cohort past the first month. Mid-market and enterprise products should be well above 85 percent, since the sales process filters out casual signups.
  • M12 retention. Healthy SMB SaaS often lands between 35 and 50 percent of the original cohort at the one-year mark. Enterprise products with annual contracts should retain 80 percent or more of accounts through the first renewal.
  • The flattening point. More important than any absolute number. Strong products flatten by month 3 to 6. If your curve is still declining at the same rate in month 9 as in month 2, no realistic acquisition budget outruns the leak.
  • Net revenue retention. Cohort NRR above 100 percent means expansion within a cohort outpaces losses, so old cohorts grow in value even as they shrink in logos. Public SaaS leaders live in the 110 to 130 percent range.

The most useful benchmark is your own history. A cohort table that improves row over row beats an impressive static number every time.

Segmented Cohorts: Where the Real Answers Live

A single cohort table tells you when churn happens. Segmented cohort tables tell you why. Split the same analysis by one variable at a time and compare curves.

By acquisition channel. Paid social cohorts retaining at half the rate of organic cohorts is one of the most common findings in SaaS, and it quietly rewrites your unit economics: a channel with cheap signups and terrible retention is often your most expensive channel per retained customer. Run this before scaling any acquisition budget.

By plan or price tier. Free-trial-to-paid cohorts vs. direct-to-paid. Monthly vs. annual billing. Starter vs. growth tier. Annual plans mechanically retain better in the first year, but watch what happens at the month-12 boundary before declaring victory.

By onboarding behavior. Cohort customers who completed a key activation step against those who did not. If customers who imported their data in week one retain at 3x the rate of those who did not, you have found your activation metric, and your onboarding flow has one job. This is the same logic behind a customer health score: find the behaviors that predict staying, then push every account toward them.

By customer size or segment. Small accounts churning fast while enterprise accounts hold is normal. The question is whether your roadmap and support model match the mix you actually retain, not the mix you acquire.

Keep segmentation honest: change one variable at a time, and stop segmenting before your cells drop below the 30-customer floor. A 12-person sub-cohort produces a beautiful chart and a meaningless conclusion.

From Table to Action: Turning Cohort Findings into Retention Fixes

A cohort table is a diagnostic, not a strategy. Here is how the common patterns map to interventions.

Finding: steep M0 to M1 drop. Customers do not reach first value. Shorten time to value aggressively: cut setup steps, pre-fill templates, add human or automated onboarding assistance for the first session. Measure the next cohort’s M1 as your success metric. Most teams see cohort-level improvement within one to two cycles because new cohorts experience the fix immediately.

Finding: slope never flattens. Value does not become habit. Push adoption of the features that correlate with long retention (your segmented table just told you which ones). Lifecycle email, in-app prompts, and quarterly reviews for higher-touch segments all aim at the same thing: moving each account from one use case to two or three.

Finding: diagonal stripe of payment churn. Fix the machinery, not the messaging. Card updaters, retry logic, pre-dunning emails before renewal charges. This is the cheapest churn you will ever recover because the customer never decided to leave.

Finding: cancellations concentrated at a lifecycle point. Intervene just before it, not after. If cohorts thin out around month 3, that is when the check-in, the pricing flexibility, or the win-back offer belongs. A cancellation flow that surfaces the right alternative at the moment of cancellation (a pause instead of a cancel, a downgrade instead of a loss, a targeted discount for price-sensitive accounts) converts a hard exit into a retained account often enough to move the cohort curve. ChurnFree’s cancellation flow and retention offers exist for exactly this point in the curve.

Finding: retention fine, revenue cohort shrinking. Downgrades, not departures. Look at plan-change data inside the cohort and at whether your packaging pushes growing customers to expand or lets them shrink quietly.

Whatever the finding, close the loop the same way: ship the fix, then watch the next two or three cohorts at the affected period. Cohorts are your experiment readout. That feedback loop, fix then verify on fresh cohorts, is the entire operating rhythm of a serious retention program, and it is what separates teams that manage churn from teams that report it. For the broader toolkit, see our guide on how to reduce churn in SaaS.

Tools for Running Retention Cohort Analysis

You have four practical options, in ascending order of effort.

Your billing or subscription platform. Stripe, Paddle, Chargebee and similar tools ship basic cohort retention charts. Zero setup, revenue-accurate, but limited segmentation.

Product analytics tools. GA4 has a cohort exploration report (Explore, then Cohort exploration) that works on activity retention. Dedicated product analytics platforms go deeper with behavioral cohorts and custom retention events. Strong for engagement questions, weaker for revenue truth, since analytics events and billing reality drift apart.

A spreadsheet plus a billing export. Underrated. A pivot table on customer start month vs. active month gets you a full logo and revenue cohort table in an afternoon, with definitions you control end to end. This is the right starting point for most teams under a few thousand customers.

Retention platforms. Purpose-built tools connect billing data to retention action. ChurnFree’s insights and analytics tie cancellation reasons, save-offer performance, and churn patterns to the same customer records, so the “why” behind a weak cohort is attached to the cohort itself. Pair the numbers with customer feedback from exit surveys and the table stops being a mystery.

Whichever tool you pick, write down your retention definition (unit, event, grain) next to the chart. Most cohort arguments inside companies turn out to be two people using two definitions.

Common Mistakes That Ruin Cohort Analysis

  • Cohorts too small to read. Below roughly 30 customers per cohort, percentages are noise. Widen the grain.
  • Mixing retention definitions. Comparing this quarter’s payment-retention table to last quarter’s activity-retention table produces confident, wrong conclusions.
  • Judging young cohorts too early. A cohort two months old tells you about onboarding, nothing else. Resist extrapolating its M12.
  • Ignoring composition shifts. If the mix of channels, plans, or customer sizes changed between cohorts, the cohort curve moved for reasons a product change cannot explain. Segment first, conclude second.
  • Confusing calendar effects with lifecycle effects. Always check the diagonal before attributing a dip to a specific cohort’s experience.
  • Reading the table and filing it. The table’s only purpose is to point the next fix at the right lifecycle moment. Track the retention metrics it feeds, but ship something.

Frequently Asked Questions

What Is Cohort Analysis?

Cohort analysis is a method that groups customers by a shared starting point, usually their signup month, and tracks each group’s behavior separately over time. Applied to retention, it shows what percentage of each signup group is still active one month, three months, and a year later, revealing patterns that a single blended retention rate hides.

How Do You Interpret a Cohort Analysis Chart?

Read it in three directions. Down a column: are newer cohorts retaining better than older ones at the same lifecycle stage? Across a row: at which month does a given cohort lose the most customers, and does its curve flatten? Along the diagonal: did something in a specific calendar month hit all cohorts at once? Column trends measure improvement, row shapes locate lifecycle problems, and diagonals expose external events.

How Do You Do a Cohort Analysis?

Define what retained means (still paying is the honest default for subscriptions), group customers by signup month, count how many from each group remain active in each subsequent month, and divide by the group’s starting size. Lay the results out as a table with cohorts as rows and months since signup as columns. A spreadsheet pivot on a billing export is enough to start.

What Is Cohort Analysis in Google Analytics?

GA4 includes a cohort exploration report under Explore. It groups users by the week or month of their first visit and tracks how many return in later periods. It measures activity retention (returning to the site or app), not payment retention, so for a subscription business it complements rather than replaces a billing-based cohort table.

What Is a Good Retention Rate for a SaaS Cohort?

It depends on segment and price point. SMB-focused SaaS commonly retains 60 to 80 percent of a cohort past month 1 and 35 to 50 percent at month 12. Enterprise SaaS with annual contracts should hold 85 percent or more through month 1 and 80 percent or more through the first renewal. The stronger signal than any absolute number is the curve’s shape: it should flatten within 3 to 6 months rather than declining indefinitely.

What Is the Difference Between Cohort Analysis and Churn Analysis?

Churn analysis measures and diagnoses customer losses overall: how many left, which segments, for what reasons. Cohort analysis adds the time dimension by tying every customer to a start date, showing when in the lifecycle losses happen and whether newer customer groups behave differently from older ones. In practice cohort analysis is the sharpest instrument inside a broader churn analysis program.

Start Reading Your Retention the Way It Actually Happens

Blended retention numbers describe your average customer, and your average customer does not exist. Cohort tables describe the customers you actually signed, month by month, and show you precisely where the leak is: the first-week cliff, the slope that never flattens, the month-12 step, the diagonal stripe of failed payments.

Build the table this week. Billing export, pivot table, thirty minutes. Read down, across, and diagonally, pick the single ugliest pattern, and aim your next retention fix at that exact point in the lifecycle. Then let the next two cohorts tell you whether it worked.

When the table points at cancellations you could have saved, that is where ChurnFree comes in: cancellation flows, targeted save offers, pause options, and the analytics to see which cohorts they rescue. Try it free and give your worst cohort pattern a better ending.