SaaS Cohort Analysis: Reading Retention Curves Right

Why Cohort Analysis Beats Aggregate Metrics

The headline MRR number lies. A SaaS company growing MRR by 8% month-over-month can have catastrophic underlying retention — the new-business growth simply masks the customer loss. Aggregate churn rates lie too: they average together cohorts with very different behavior, hiding the fact that recent cohorts may be dramatically worse (or better) than older ones.

Cohort analysis fixes this. By grouping customers by their signup or first-purchase month and tracking each group over time, you see retention quality directly. The cohort matrix is the most diagnostically useful artifact in SaaS finance — more revealing than any single dashboard metric. This guide walks through building one, reading it, and acting on the patterns. It assumes you already understand the basics of SaaS metrics (MRR, ARR, NRR, GRR) and want to go one level deeper.

What Is a Cohort in SaaS?

A cohort is a group of customers defined by a shared starting event. The most common cohort definition: month of first paid transaction. Other useful cohort definitions:

Cohort Type	When to Use
Acquisition month	Default — tracks retention from first paid event
Trial start month	Useful when conversion path matters
Acquisition channel	Compares retention across paid vs organic vs referral
Plan tier at signup	Compares retention across starter / professional / enterprise
Industry / segment	Reveals which customer segments retain best
Pricing version	When you change pricing, separate before/after cohorts

Most companies start with acquisition-month cohorts. Once that view is stable, add channel and tier overlays.

The Cohort Matrix: How to Build One

The cohort matrix is a table where rows are cohorts (signup months) and columns are months of tenure. Each cell shows retention for that cohort at that tenure point.

Here's a worked example for a B2B SaaS startup tracking logo retention (the % of original customers still paying):

Cohort	M0	M1	M2	M3	M6	M12
Jan 2025 (40 customers)	100%	88%	78%	73%	64%	58%
Feb 2025 (52 customers)	100%	92%	85%	80%	71%	64%
Mar 2025 (61 customers)	100%	95%	88%	84%	76%	—
Apr 2025 (74 customers)	100%	94%	89%	85%	—	—
May 2025 (88 customers)	100%	96%	91%	—	—	—

Reading down a column shows whether each successive cohort is improving (a positive sign — your product, onboarding, or targeting is getting better). Reading across a row shows the retention curve for a single cohort over time.

The Jan 2025 cohort shows 58% retention at month 12 — meaning 42% of the original 40 customers churned within their first year. The Mar 2025 cohort retains 76% at month 6, compared to Jan's 64% at the same tenure — a meaningful improvement. The company is getting better at retention.

Tools That Build the Matrix

Spreadsheet (Google Sheets, Excel): perfectly fine for early-stage companies. The Andrew Chen cohort spreadsheet is the standard starting template.
Stripe Sigma / ChartMogul / ProfitWell: built-in cohort views for SaaS billing data
Amplitude / Mixpanel: behavioral cohort analysis for product retention (not just billing)
Custom SQL: most flexible; required once you have multiple cohort dimensions

For companies under $1M ARR, a spreadsheet is sufficient. Adding tooling earlier is over-investment.

Three Curve Shapes and What They Mean

Every cohort produces a retention curve. The shape — not the average — reveals product-market fit.

The Smile Curve (Strong PMF)

The curve drops sharply in months 1–3, then stabilizes or even rises slightly. The flat (or upward) tail is the critical signal: you've reached a stable retained base. Customers who survive the early churn period stay for years.

What it looks like: 100% → 75% → 70% → 68% → 68% → 70% → 71%

This is the shape healthy B2B SaaS companies see. The early drop is normal (some customers always sign up and discover they don't need the product). The flat tail proves the surviving customers have found enduring value.

The Slope Curve (Leaky Bucket)

The curve declines steadily, with no flattening. Each month loses 4–8% of remaining customers, indefinitely.

What it looks like: 100% → 88% → 78% → 70% → 62% → 56% → 50% → 44%

This shape is fatal at scale. You're acquiring customers as fast as you're losing them — running up a down escalator. The fix is not better acquisition. The fix is product-side: improved onboarding, better activation, addressing the feedback loop gap that causes consistent dissatisfaction.

The Cliff Curve (Onboarding or Trial-Mismatch Problem)

The curve drops dramatically in months 0–2 (say, 100% → 50% → 38%), then flattens.

What it looks like: 100% → 55% → 42% → 38% → 36% → 35%

The cliff signals an early-stage mismatch — typically poor onboarding, trial-to-paid users who didn't actually want the product, or activation flow that fails to deliver value. The base that survives often retains well; the problem is too few customers reach the stable base.

The fix here is upstream: onboarding redesign, better-qualified trial signups, and pricing changes that filter unfit customers before they convert.

Logo Retention vs Dollar Retention: Why Both Matter

The cohort matrix above tracks logo retention (% of customers still paying). Dollar retention (% of original MRR still being collected) is often dramatically different.

Example: A cohort starts with 100 customers paying $100/month average ($10,000 MRR). 12 months later:

60 of the original customers remain (logo retention: 60%)
Those 60 customers now pay $180/month average (some upgraded, a few downgraded)
Cohort revenue: 60 × $180 = $10,800
Dollar retention: 108%

This cohort shows 60% logo retention but 108% dollar retention. The losing-and-expanding pattern is structurally common: the churned customers were typically the smallest accounts, and the surviving customers tend to expand. This is healthy.

The diagnostic gap to watch: when dollar retention is high while logo retention is poor, you're acquiring fundamentally unfit customers and expanding the few that fit. This works at small scale but can't scale forever — you'll eventually exhaust the supply of fit customers.

Cohort Diagnostics: What Investors Look For

When investors review cohort data, they look for specific patterns:

Pattern	What It Signals
Each newer cohort retains better than older ones	Product/process improving — green flag
Cohorts are stable cohort-over-cohort	Mature, predictable business — neutral
Recent cohorts retain worse than older ones	Something broke — investigate before raising
Month-12 logo retention 80%+ for SMB SaaS	Healthy
Month-12 logo retention 70%+ for mid-market	Healthy
Month-12 logo retention 90%+ for enterprise	Healthy
Flat retention tail (months 12+)	Found durable customer fit
Continuously declining retention	Leaky bucket — fix product before scaling acquisition
Wide variance across cohorts	Inconsistent acquisition quality

The "flat tail" is the single signal investors care about most. A retention curve that doesn't flatten is a business that doesn't have a defensible base.

Acting on Cohort Insights

The patterns in your cohort matrix should drive specific actions:

When Recent Cohorts Underperform

Investigate channel mix (did you shift to a worse-retaining channel?)
Review pricing changes (did a discount campaign attract unfit buyers?)
Audit onboarding (did a recent change degrade activation?)
Check sales motion (are reps closing deals that shouldn't close?)

When Specific Tenure Drops Are Sharp

Month 1 drop: onboarding failure — see our user onboarding design playbook
Month 3–4 drop: post-honeymoon disillusionment — gap between sales promise and product
Month 11–13 drop: annual contract renewal failures — assess CS motion
Sudden drop in a specific month across all cohorts: external event (outage, competitor launch, price change)

When Logo vs Dollar Retention Diverge

Strong expansion + poor logo retention: acquiring the wrong customers; tighten targeting
Stable logo + poor dollar retention: customers are downgrading; investigate why
Both healthy: stable business that can scale

A Worked Diagnostic Walkthrough

A SaaS startup at $500K ARR reports the following cohort behavior:

Jan–Mar 2025 cohorts: 65–70% month-12 retention
Apr–Jun 2025 cohorts: 55–62% month-12 retention (declining)
Jul–Sep 2025 cohorts: 50–55% projected at month 12 (still developing)
Aggregate logo churn: 4.2% monthly (up from 2.8% a year ago)
Dollar retention has remained 105–110% throughout

The diagnosis: Logo retention is degrading across recent cohorts. Dollar retention is stable because surviving customers expand. But the long-run dynamic is unsustainable — the company is acquiring an increasing share of unfit customers, masking the problem with expansion revenue from a shrinking core.

The action: Audit acquisition channels for Q2 and Q3 — what changed? Possible causes: a paid campaign attracting price-sensitive buyers, a partner channel sending mismatched leads, an SEO change driving traffic with different intent. Pause or fix the channel mix before adding more acquisition spend.

This is the type of diagnosis aggregate metrics can't surface. Cohort analysis makes it visible.

When Cohort Analysis Doesn't Apply (Not For You)

Skip detailed cohort analysis if:

You have fewer than 100 paying customers. Cohort sizes are too small for statistical signal. Aggregate metrics plus customer interviews are higher-ROI.
Your business is transactional, not subscription. E-commerce cohorts work differently (repeat-purchase rate, time-to-second-order) — the SaaS framework doesn't apply directly.
You're under 12 months old. You don't have enough cohorts to compare trajectories. Track the data anyway; analysis matures over time.
You've changed pricing, plans, or product radically. Pre-change and post-change cohorts aren't comparable — segment them and avoid drawing conclusions across the boundary.

Conclusion

Cohort analysis is the single most useful diagnostic in SaaS finance. The cohort matrix surfaces what aggregate metrics hide: whether your retention is improving over time, where in the customer lifecycle you're losing people, and whether your expansion revenue is masking a leaky bucket.

Build the matrix in a spreadsheet. Track both logo and dollar retention. Watch the shape of the curve, not just the average. Pair this with the broader SaaS metrics framework, disciplined customer feedback loops, and a clear user onboarding design — together they form the operating system that separates SaaS companies that compound from those that just grow.

Frequently Asked Questions

What is cohort analysis in SaaS?

Cohort analysis groups customers by a shared starting event (typically their signup month) and tracks how each group's retention and revenue behave over time. Unlike aggregate metrics (which average all customers together), cohort analysis isolates the behavior of each group — showing whether retention is improving, stable, or declining across newer cohorts.

What's the difference between logo retention and dollar retention?

Logo retention is the percentage of original customers still paying. Dollar retention is the percentage of original MRR still being collected. They often diverge: a cohort can have 60% logo retention but 110% dollar retention if surviving customers expanded. Track both — they reveal different aspects of customer health.

How long should I run cohort analysis?

You need at least 12 months of cohorts to see the full retention curve shape. Many businesses track to 24 or 36 months to confirm the 'flat tail' of long-term retention. Companies under a year old can still build the matrix, but recognize the curves are incomplete and trajectories aren't yet meaningful.

What's a healthy month-12 retention rate for SaaS?

80%+ logo retention for SMB SaaS, 85%+ for mid-market, 90%+ for enterprise. Dollar retention should be 100%+ at month 12 for SMB (some upselling) and 110%+ for mid-market and enterprise. Below these floors, you have a retention problem that capital can't fix — only product and onboarding work will.

What tool should I use for cohort analysis?

For companies under $1M ARR, a spreadsheet is sufficient — Andrew Chen's free template is the standard. As you scale, dedicated tools (ChartMogul, ProfitWell, Maxio) provide automation and richer visualizations. Behavioral cohort analysis (vs billing-only) requires Amplitude or Mixpanel. Don't add tools before you've outgrown the spreadsheet.

Why does my recent cohort retain worse than older ones?

Three common causes: (1) channel mix shifted — you started acquiring a less-fit customer segment. (2) Pricing or packaging change attracted different buyers. (3) Product or onboarding regression. Investigate channel-by-channel attribution before assuming the product is at fault.

What's the 'flat tail' in cohort analysis?

The point in the retention curve where it stops declining and stabilizes — typically 12–24 months in for healthy SaaS. A flat tail means you've found a durable customer base that retains long-term. A continuously declining tail means you haven't yet — your business is a leaky bucket, not a stable subscription. Investors look for flat tails before pricing companies confidently.