What Is Churn Rate?: Customer, revenue, and cohort formulas with a worked example

Churn rate is the share of a defined customer, account, subscriber, or recurring-revenue base lost during a stated period. Customer churn asks how many entities left. Revenue churn asks how much recurring value disappeared. Cohort churn asks what happened to one fixed group over comparable lifecycle time. A churn percentage without its unit, denominator, loss event, and window is not a complete metric.

Stripe’s billing documentation separates subscriber and revenue views and defines churned revenue as churned monthly recurring revenue plus contraction monthly recurring revenue. Its cohort reporting instead compares the recurring revenue remaining from a fixed cohort with that cohort’s starting revenue. Those are related measures, not interchangeable labels.

There is no universal churn benchmark in the source set. Contract length, customer segment, business model, counting unit, revenue definition, and observation window all change the result. The defensible first comparison is with the same metric under the same contract over time.

The four churn measures answer different questions#

The table is a measurement design, not an accounting standard. A team still has to define what counts as a customer, when a cancellation becomes effective, which revenue is recurring, and how refunds, migrations, pauses, reactivations, downgrades, currencies, and acquisitions are handled.

Customer churn rate#

A common customer-churn contract is:

Customer churn rate = customers lost during the period
                      ÷ customers present at the start of the period
                      × 100

Using the starting population prevents new customers acquired during the period from diluting losses in that original base. Some vendor dashboards use other conventions, so exported numbers should be reconciled to the displayed definition before comparison.

Customer churn is also called logo churn when the counted entity is a customer account. In a self-serve product, the entity might be a paid subscription instead. These units are not automatically equal: one account can hold several subscriptions, and one buyer can represent multiple users.

Gross and net revenue churn#

Gross revenue churn isolates recurring revenue lost from the starting base:

Gross revenue churn rate = (churned recurring revenue + contraction recurring revenue)
                           ÷ starting recurring revenue
                           × 100

Net revenue churn offsets those losses with expansion from customers retained from the starting base:

Net revenue churn rate = (churned recurring revenue + contraction recurring revenue
                          - expansion recurring revenue)
                         ÷ starting recurring revenue
                         × 100

Stripe documents that revenue retention can exceed 100% when expansion is greater than lost revenue. Under the matching contract, net revenue churn would then be negative. That does not mean no customers left; it means expansion from retained customers more than offset the recurring-revenue loss.

Cohort churn#

A cohort fixes the population at a shared starting event, such as first paid month, activation month, or contract start. It then compares each cohort at the same lifecycle age.

Cohort revenue retention at month n = cohort recurring revenue remaining at month n
                                      ÷ cohort recurring revenue at start
                                      × 100

Cohort revenue churn at month n = 100% - cohort revenue retention at month n

The subtraction is valid only when retention and churn use the same cohort, value unit, window, and treatment of expansion. If the retention view permits expansion above the starting base, cohort churn can become negative. A fixed customer-count cohort behaves differently because an entity count cannot be increased by spending more.

All-base monthly churn and cohort churn answer different questions. The all-base view tracks the current book of business. The cohort view controls for lifecycle age and can reveal whether newer cohorts behave differently from older ones.

Worked example: one month, three answers#

The following figures are synthetic arithmetic, not company data or a benchmark.

An account base starts the month with 200 customers and $40,000 in monthly recurring revenue. During the month:

• 10 starting customers leave, removing $2,400;
• retained customers contract by $600; and
• other retained customers expand by $1,400.

Customer churn = 10 ÷ 200 = 5%

Gross revenue churn = ($2,400 + $600) ÷ $40,000 = 7.5%

Net revenue churn = ($2,400 + $600 - $1,400) ÷ $40,000 = 4%

The 5% result says how many starting customers left. The 7.5% result says how much starting recurring revenue was lost before expansion. The 4% result says how much loss remained after expansion. None is the single correct churn rate; each is correct only for its declared question.

Suppose the same 200 customers form a fixed cohort. If $37,000 of their original recurring revenue remains at month end before counting expansion, gross cohort revenue retention is 92.5% and gross cohort churn is 7.5%. That agreement is a reconciliation check because the example uses the same base and gross-loss rules.

Build a churn metric contract before a dashboard#

Churn and retention are not always complements#

Customer churn and customer retention can sum to 100% when they partition the same starting population into lost and retained entities. They stop being complements when teams mix definitions—for example, period churn over the entire active base versus retention of a newly acquired cohort.

Revenue measures require still more care. Gross revenue retention and gross revenue churn can be complements under matching rules. Net revenue retention can exceed 100% because it includes expansion; its corresponding net churn can be below zero. Stripe’s subscription analytics documentation makes this possibility explicit.

What churn cannot tell you#

Churn identifies a measured loss. It does not prove why the loss happened. Plan, segment, tenure, product behavior, support history, billing events, and interviews can narrow hypotheses, but an association is not a causal explanation.

Avoid labeling every downgrade, inactive user, failed charge, or non-renewal as the same event. The operational response differs. A product adoption problem, a failed payment, a planned seasonal pause, and an account lost to a competitor need different owners and evidence.