Churn rate is computed as subscribers lost in period / subscribers active at start of period, and the period has to be stated because monthly and annual figures are not interchangeable. A 5% monthly churn rate compounds to roughly 46% annual churn (1 - (1 - 0.05)^12 ≈ 0.4596), not 60%. Annualizing by multiplying overstates the loss and distorts every downstream LTV estimate the number feeds.
Two distinctions change the read. Gross churn counts cancellations only; net churn subtracts winbacks and upgrades within the same period, and can go negative for SaaS-style products with expansion revenue but rarely for consumer subscriptions. Voluntary churn is active cancellation; involuntary churn is payment failure such as expired cards or declined transactions. Operators commonly find involuntary churn to be a non-trivial slice of total cancellations, with the size varying by category, price point, and payment mix. The bulk of it is addressable with card-updater services and dunning flows.
Cohorts read more honestly than aggregates. New-cohort churn in months 1–3 reflects product-market fit being tested; steady-state churn in month 6 and beyond reflects how the surviving base behaves. Reading them as one number averages a transient over a steady state. Pause is not churn, but paused subscribers are widely reported to cancel at a higher rate than active ones over the following months, so paused accounts belong in the denominator as a leading indicator rather than as inert inventory.
Patterns to watch for: “active subscriber” counts that quietly include paused or skipped accounts, periods that exclude the early-month shedding, gross or net reported whichever flatters. State the period, state the definition, separate cohorts.