Cohort Analysis for SaaS Retention: Definition & Examples (2026)
Cohort analysis is a retention measurement technique that groups customers by a shared starting event — usually their sign-up month — and tracks how each group's retention or revenue evolves over time. It is the only way to tell whether retention is actually improving or just looks better because of growth mix.
How a cohort table works
Each row is a cohort (the customers who signed up in a given month). Each column is the number of months elapsed since sign-up. The cell shows the percentage of that cohort still active.
| Cohort | M0 | M1 | M2 | M3 |
|---|---|---|---|---|
| January | 100% | 85% | 72% | 65% |
| February | 100% | 88% | 76% | 70% |
| March | 100% | 91% | 82% | — |
In the table above, each newer cohort is retaining better at every elapsed month — a clear signal that onboarding or PMF improvements landed somewhere between January and March. A blended monthly churn number would have hidden this entirely.
The cohort math
Retention(N) = (active customers at month N) ÷ (customers in original cohort) × 100
For revenue cohorts, swap "active customers" for "MRR from surviving customers in the cohort." Revenue cohorts can exceed 100% when expansion is healthy — that is the visual signature of negative net churn.
What cohort analysis reveals
- Onboarding-quality changes. Cohorts after an onboarding rebuild should retain better at M1 and M2 than earlier cohorts.
- Product-market fit improvements. A flattening retention curve across newer cohorts signals deeper fit.
- ICP shifts. If new cohorts retain worse than old ones, you are likely selling to the wrong segment.
- Pricing-page tests. Cohort retention by plan tier reveals whether new packaging attracts stickier customers.
Common cohort analysis mistakes
- Mixing trial and paid customers. They behave completely differently — separate cohorts.
- Reading immature cohorts. A March cohort cannot have an M6 number in April. Show only mature cells.
- Small cohorts. Below 30 customers per cohort, single-customer churn moves the percentage by >3pp. Aggregate to quarterly cohorts instead.
- Ignoring segments. SMB and mid-market cohort curves are wildly different. Always cohort within segment.
Cohort analysis + behavioral churn signals
Cohort analysis is descriptive — it tells you what happened to a group over time. Behavioral churn prediction is forward-looking — it tells you which individual accounts are losing engagement right now. The two together cover the full retention lifecycle: cohort tables guide where to invest at the program level, behavioral scoring guides where CSMs spend the next hour. For deeper reading, see our 2026 SaaS churn rate benchmarks.
Frequently asked questions
What is cohort analysis?
Cohort analysis is a retention measurement technique that groups customers by a shared starting event — usually the sign-up month — and tracks how each group's retention or revenue evolves over subsequent months. Unlike a single blended churn number, cohort tables show whether retention is improving or decaying over time.
Why is cohort analysis better than a single churn number?
A blended monthly churn number averages new and tenured customers together, hiding both early-tenure churn spikes and improvements in onboarding quality. Cohort analysis separates those signals so product, success, and finance can act on the right one.
What does a healthy SaaS retention curve look like?
A healthy B2B SaaS retention curve flattens after month 3–6 — the steep drop of early-tenure churn ends and the surviving cohort stabilizes. The flatter the asymptote, the stronger the product-market fit. Curves that keep decaying linearly indicate weak fit or a value-mismatch problem.
How many cohorts do I need to draw conclusions?
For directional reads, three consecutive monthly cohorts with at least 30 customers each. For statistical confidence — especially when comparing onboarding changes — six cohorts with 100+ customers each. Below that, run a holdout experiment instead.