SaaS Churn Rate: Customers, Revenue, and Retention Explained

Begin with a defined starting cohort

A monthly customer churn rate usually compares customers lost during the month with customers active at the start. In this guide, newly acquired customers are not added to that starting denominator.

Customer churn = customers lost from the opening cohort ÷ opening customers

If a month begins with 500 customers and 20 of those customers leave, churn is 4%. Adding 50 new customers during the month does not change this cohort's 4% churn rate. It changes the ending customer count.

Write down how the business treats trial users, overdue accounts, pauses, reactivations, and multiple subscriptions belonging to one customer. Billing systems can use configurable definitions, so exports should be reconciled before figures are compared. ^[1]

Revenue churn asks a different question

Losing ten small accounts is not equivalent to losing ten enterprise accounts. Start with monthly recurring revenue, or MRR, belonging to the opening customer cohort. Identify the amounts lost through cancellation and contraction.

Gross revenue churn = (churned MRR + contraction MRR) ÷ opening MRR

Gross revenue retention = 1 − gross revenue churn

Net revenue retention = (opening MRR − churn − contraction + expansion) ÷ opening MRR

This guide treats expansion from existing opening-cohort customers separately from revenue from new customers. Keep reactivation treatment explicit rather than automatically putting it into whichever category improves the result.

Worked example: customers leave while revenue expands

Assume opening MRR of $50,000. During the month, cancellations remove $2,000, downgrades remove $1,000, and upgrades within the opening cohort add $4,000.

Negative net revenue churn means expansion more than offset the modeled losses. It does not mean no customers cancelled. New-customer revenue is excluded from all four calculations in this table.

Monthly churn does not annualize by simple multiplication

If a starting cohort loses a constant 4% each month and there are no reactivations, its modeled twelve-month retention is 0.96¹², approximately 61.27%. Its cumulative churn is therefore approximately 38.73%, not 48%.

This is a mathematical projection under a constant-rate assumption, not a forecast of actual customers. Annual contracts often produce renewal clusters, and customer risk can change with tenure. A single month is rarely enough evidence to assume a stable long-run rate.

Do not substitute net revenue churn into a customer-lifetime formula. Revenue expansion and individual customer survival are different processes.

Investigate churn by cohort and reason

Break the top-line result into segments that can support a decision: acquisition cohort, plan, customer size, use case, or renewal type. Keep groups large enough that individual accounts do not create misleading swings.

Separate voluntary cancellation from payment failure when that distinction can guide action. Then look for underlying evidence: onboarding completion, product usage, support issues, renewal conversations, and whether the product delivered the expected outcome.

Cancellation surveys can be useful but are not a complete explanation. People who respond may differ from those who do not. Combine their responses with observed customer behavior rather than treating a short label as a proven cause.

Build a reproducible retention report

Store the reporting window, opening cohort definition, currency treatment, discount policy, and cancellation timing with each report. Reconcile the opening and ending balances using a roll-forward that includes new, expanded, contracted, churned, and reactivated revenue where applicable.

A good retention dashboard should make an uncomfortable result visible, not optimize the definition until the result looks better. Consistency over time is more valuable than a favorable comparison to an unrelated business.

Use this guide thoughtfully. Examples illustrate a calculation method, not a guaranteed outcome. The usefulness of any result depends on the definitions, measurements, and assumptions used.