SaaS cohort analysis: how a buyer reads your cohorts

SaaS cohort analysis groups customers by the month (or quarter, or year) they started paying. It then tracks what share of each group, and of each group’s revenue, is still there as time passes. A buyer cares because a cohort table separates two things that a single retention rate blends together: how good your retention used to be, and how good it is for the customers you’re signing now.

This guide covers how cohorts are built from billing data, the two kinds of cohort table, and the patterns that lead an acquirer to ask follow-up questions.

What a cohort is in a subscription business

A cohort is every customer who started paying in the same period. “Started paying” needs a precise rule, because the rule decides which row a customer lands in:

• A customer joins a cohort at the start of the service period of their first recurring charge with a positive amount, once that invoice is paid.
• Void, draft and uncollectible invoices don’t start a cohort. A customer whose first invoice was never collected hasn’t really started.
• Free trials don’t count until the first paid period begins.

Get this wrong and cohorts drift: customers land in the month they signed up for a trial, or in a month where they never paid.

Logo retention and revenue retention

There are two cohort tables, and a buyer will want both.

• Logo (customer) retention: of the customers who started in a period, how many still pay n months later. It usually falls over time, and rises only when customers who left come back.
• Revenue retention: of the MRR those customers started with, how much they pay now. It includes expansion, so a cohort can go above 100%.

A revenue cohort above 100% looks great. It can also hide a cohort that has lost half its customers, with the survivors paying more. That is why buyers ask for the logo table next to the revenue table.

Reading the table

Here’s a revenue-retention table by signup quarter. These are illustrative numbers, not data from a real company.

Cohort	Customers	Starting MRR	Month 3	Month 6	Month 12
Jan 2025	40	$4,000	96%	94%	101%
Apr 2025	52	$5,200	93%	88%	90%
Jul 2025	61	$6,100	89%	82%	—
Oct 2025	70	$7,000	85%	77%	—

As of June 2026. A dash means the cohort isn’t old enough yet.

There are three ways to read a table like this:

• Across a row: how one cohort behaves over its life. The January 2025 cohort dipped, then expanded past where it started.
• Down a column: whether newer cohorts retain as well as older ones at the same age. Here, month-6 retention falls from 94% to 77% across four cohorts. That’s the pattern a buyer will ask about first.
• Along a diagonal: cells that fall in the same calendar month. If every cohort dips in the same month, look for a calendar event, such as a price change, an outage or a billing migration, before blaming customer quality.

Patterns that prompt diligence questions

• Newer cohorts decaying faster. As in the example. The questions that follow are about what changed in acquisition: a new channel, a cheaper plan, a looser trial, a different customer segment.
• Revenue retention above 100% with weak logo retention. Expansion from a few customers is covering for many who left. This connects directly to customer concentration.
• One cohort that looks unusually strong. Often a single large customer. An analyst will look at that cohort’s customer list.
• Cohorts too small to mean anything. Five customers in a month makes every percentage noisy. Quarterly or yearly cohorts are more honest for a small customer base.

Cohorts by first plan

A useful cut is to group customers by the plan they first bought, not just by when they joined. It answers questions like:

• Did customers who came in on the cheaper plan you launched last year retain as well as the ones on your core plan?
• Does the annual plan retain better than monthly, or does it just delay churn to the renewal date?
• Is one plan’s poor retention dragging down the average?

If retention differs a lot by entry plan, a buyer will discount the average and look at the plan mix going forward. It’s better to know that before they do.

Monthly, quarterly or yearly cohorts

• Monthly cohorts show the most detail and need enough customers per month to be readable.
• Quarterly cohorts are the usual compromise for diligence.
• Yearly cohorts, read on signup anniversaries, are the clearest long-term view and work even for a small business.

Show the grain your customer count supports, and say which one you’re showing.

How Bright Vault shows this

Bright Vault builds signup cohorts from your Stripe history by month, quarter and year, using the membership rule above. It shows both customer and revenue retention and groups cohorts by first plan. Yearly tables are read on signup anniversaries. Every view can be filtered by plan. Reports are private to your team.

Cohorts explain who stays. To see how that adds up across the whole customer base, read how a buyer reads net revenue retention. For where cohorts sit among a buyer’s other requests, see the due diligence checklist.

FAQ

What’s the difference between cohort retention and NRR?

Cohort retention follows customers grouped by start date over their whole life. Net revenue retention takes everyone who was a customer at the start of a period, often the last twelve months, and measures what they pay at the end of it. NRR is one number. A cohort table shows how that number is built.

How many cohorts do I need to show?

Enough to see a trend at the same age. For quarterly cohorts, two years of signups gives eight rows, which is usually enough to show whether newer customers retain like older ones.

Should cohorts use signup date or first payment date?

First payment, or more precisely, the start of the first paid service period. Signup and trial dates put customers into cohorts before they’ve paid.