SaaS Cohort Analysis: How to Track Retention by Signup Month

A single churn rate tells you that customers are leaving. It doesn't tell you whether it's your newest customers leaving fast or your oldest customers finally giving up, whether a pricing change last quarter quietly broke retention, or whether things are actually getting better. Cohort analysis answers all three by grouping customers by when they signed up and tracking each group separately over time.

This post covers what cohort analysis is, how to read a cohort table, the difference between logo and revenue cohorts, and what good retention curves actually look like.

What Is Cohort Analysis?

Cohort analysis groups customers by a shared starting point - usually their signup month - and tracks what percentage of each group is still active (or still paying) in each month afterward.

🧒 Explained simply Picture every customer who started buying lemonade in June as one group, and everyone who started in July as a separate group. Instead of asking "how many customers do I have total," you ask "out of everyone who joined in June, how many are still buying lemonade in July, August, September?" Tracking each starting group separately shows you exactly when and why people stop coming back.

Rather than one churn number for the whole business, you get a separate retention curve for every cohort - which makes it possible to see whether retention is improving, worsening, or stable as your product, pricing, or onboarding changes over time.

How to Read a Cohort Retention Table

A cohort table is typically laid out as a triangle: each row is a signup cohort, each column is months since signup, and each cell shows the percentage of that cohort still active.

Cohort	Month 0	Month 1	Month 2	Month 3	Month 6	Month 12
Jan signups	100%	91%	85%	81%	74%	68%
Feb signups	100%	93%	87%	84%	77%	-
Mar signups	100%	95%	90%	87%	-	-

Reading this table tells a story a single churn number can't: retention at every stage is improving cohort over cohort (95% vs 93% vs 91% at Month 1), which suggests something - onboarding, pricing, ICP targeting - genuinely got better between January and March.

A flattening curve is a good sign. If retention drops sharply early and then levels off (a "hockey stick" shape), it means you've found a stable core of customers who get lasting value - the early drop-off is unhappy-fit customers self-selecting out.

A curve that keeps sloping toward zero is a leaky bucket. If retention never flattens and keeps declining steadily out to month 12 and beyond, there's no stable retained base - you're rebuilding your customer base from scratch every year, no matter how much new signup volume you generate.

Types of Cohorts Worth Tracking

Signup-month cohorts are the default and the most common - they isolate the effect of time-based changes like a pricing update, a new onboarding flow, or a different marketing channel mix.

Acquisition-channel cohorts group customers by how they found you (paid ads, organic, referral, partner). Channels often have very different retention profiles even at similar acquisition cost, which channel-blind reporting hides completely.

Plan-based cohorts group customers by which plan they signed up on. It's common to find that a cheaper entry plan has structurally worse retention than a mid-tier plan - useful information for pricing and packaging decisions.

Behavior-based cohorts group customers by an early action (completed onboarding within 24 hours vs not, used a specific feature in week one vs not). These cohorts are the most useful for finding causal activation levers, since they isolate behavior rather than just timing.

Logo Cohorts vs Revenue Cohorts

Cohort analysis can track either the percentage of customers retained (logo cohort) or the percentage of revenue retained (revenue cohort) - and the two frequently diverge.

	Logo Cohort Retention	Revenue Cohort Retention
Measures	% of customers still active	% of original MRR still being paid
Can exceed 100%?	No - never gains customers back	Yes - if expansion outpaces churn within the cohort
Best for	Understanding product stickiness	Understanding net revenue dynamics

A revenue cohort retention curve that rises above 100% over time is the cohort-level signal of net negative churn - the surviving customers in that cohort are expanding enough to more than offset the ones who left. This is one of the strongest health signals a SaaS business can show, and it's invisible if you only ever look at logo retention.

What Good Cohort Retention Looks Like

Benchmarks vary by segment, but rough patterns hold across most B2B SaaS:

Stage	Healthy Month 1 Retention	Healthy Month 12 Retention (flattened)
SMB / self-serve	80-90%	40-60%
Mid-market	90-95%	65-80%
Enterprise	95%+	85%+

What matters more than hitting an exact number is whether the curve flattens at all, and whether each newer cohort retains better than the one before it. A business with a 70% Month 12 floor that's stable and improving cohort over cohort is in a stronger position than one with an 85% number that's quietly eroding.

Common Cohort Analysis Mistakes

Mixing logo and revenue cohorts in the same chart. They answer different questions and should be analyzed separately - combining them in one view tends to produce a misleading composite number.

Drawing conclusions from a window that's too short. A cohort only three months old can't tell you whether its curve will flatten or keep sloping down. Early read-outs are directional at best.

Ignoring cohort size. A cohort of 8 customers from a slow signup month will show noisy, jumpy retention percentages that look dramatic but reflect small-sample randomness, not a real trend.

Comparing cohorts acquired through very different channels or campaigns as if they were equivalent. A cohort acquired through a heavily discounted promotion will almost always retain worse than an organic cohort from the same month - the comparison needs that context to be useful.

How to Track Cohort Retention in Chartsy

Chartsy builds cohort retention tables directly from your Stripe or Paddle subscription data, broken down by signup month, plan, or acquisition source. You can ask:

"Show me cohort retention by signup month for the last 12 months"
"Compare retention between customers on the Starter plan vs Growth plan"
"What's my revenue cohort retention curve for customers who signed up in Q1?"
"Which signup month had the best 6-month retention?"

Connect Stripe and see your cohort retention curves →

Frequently Asked Questions About SaaS Cohort Analysis

What is cohort analysis in SaaS? Cohort analysis groups customers by a shared starting point - typically signup month - and tracks what percentage of each group remains active or paying over time. It reveals retention patterns that a single aggregate churn number hides, including whether retention is improving or worsening as your product and onboarding change.

What's the difference between logo cohorts and revenue cohorts? Logo cohorts track the percentage of customers still active over time. Revenue cohorts track the percentage of original MRR still being paid. Revenue cohorts can exceed 100% if expansion revenue from surviving customers outpaces what was lost to churn - logo cohorts never can, since they only ever lose customers.

What does a flattening retention curve mean? A curve that drops in the early months and then levels off indicates you've found a stable core of customers getting lasting value - the early drop-off represents poor-fit customers self-selecting out. A curve that never flattens and keeps declining means there's no durable retained base, regardless of how much new signup volume you bring in.

How many months of data do I need before trusting a cohort? At least 3-6 months to see an early trend, and ideally 12 months to see whether the curve flattens. Cohorts younger than that can only tell you the early trajectory, not whether retention will stabilize.

Why does my newest cohort look worse than older ones? This can mean a real regression (a recent pricing, onboarding, or targeting change hurt retention) or simply that the cohort hasn't had time to show its later-stage flattening yet. Compare it to how older cohorts looked at the same number of months post-signup, not to their current, more mature numbers.