Here is a small honesty test for a familiar metric. Take last month's churn rate — the number in the dashboard, the one quoted in the meeting. Now ask a simple follow-up: how many of those "churned" subscribers actually decided to leave? In most subscription businesses, a surprising share never decided anything. Their card expired, their bank declined a charge, a retry schedule ran out, and the system closed the account. The formula counted them exactly the same as a customer who clicked cancel in frustration. Those are different problems, and adding them together produces a number that misleads everyone who acts on it.
The standard formula, and what it hides
The textbook calculation is not wrong so much as incomplete:
Churn rate = subscribers lost during the period ÷ subscribers at the start of the period
Run it, by all means. Just do not stop there, because the numerator is a blend. It mixes voluntary churn — a person choosing to end the relationship — with involuntary churn, where the relationship ended because a payment failed and was never recovered. Product, pricing, and onboarding work move the first kind. Billing operations, retry timing, and card-update flows move the second. A single blended rate tells a team something went wrong without telling them whose job the fix is.
A worked example with 1,000 subscribers
Start the month with 1,000 subscribers. During the month, 50 subscriptions end, so the standard calculation reports:
50 ÷ 1,000 = 5.0% churn
Now open the exit records and split them honestly. Of the 50 endings, 32 were voluntary cancellations: the subscriber cancelled, confirmed, and left. The other 18 were involuntary: their payments entered past-due, the available retries and update requests did not recover them inside the recovery window, and the account was closed for non-payment. The split calculation is just as simple:
- Voluntary churn: 32 ÷ 1,000 = 3.2%
- Involuntary churn: 18 ÷ 1,000 = 1.8%
- Total: 3.2% + 1.8% = 5.0%
Nothing about the total changes. Everything about the response changes. At 5.0% blended, a team might commission exit surveys, redesign onboarding, or discount renewals. At 3.2% voluntary with a 1.8-point billing leak, the highest-return project is probably in the payment flow — and a discount offered to someone whose card expired would not have saved them anyway.
One more number belongs in the worksheet. Those 18 involuntary exits were the residue of a larger group: say 47 subscribers entered past-due during the month, and 29 recovered and stayed. Count only people whose accounts actually closed as involuntary churn. The 29 who paid are not churn at any rate — they are recovered revenue, and counting an early failed attempt as an exit is how people who never left end up inflating the metric.
How to do the split in practice
Step 1: Fix your denominator. Use subscribers at the start of the period, and use it consistently. Some teams prefer an average of start and end counts; either is defensible, switching between them month to month is not.
Step 2: Tag every exit with a reason at close, not later. Voluntary means the subscriber initiated the ending. Involuntary means the account closed after a failed payment exhausted recovery. Do this when the account closes, while the billing record is still attached to it.
Step 3: Close the recovery window before you count. Pick a defined window — for example, until the final retry and update request have run their course — and only count involuntary exits after it ends. Accounts still in recovery are past-due, not churned.
Step 4: Report three lines, every time. Total, voluntary, and involuntary, side by side, with the past-due recovery count next to them. If you want the full worksheet version of how to calculate churn rate with these denominators and windows spelled out, start there and adapt it to your billing cycle.
Why the split changes decisions
Budgets follow the split. A voluntary problem argues for product work, clearer value, and save offers at the cancel moment. An involuntary problem argues for earlier card-expiry warnings, retries timed to when funds are actually likely to be present, and update pages a tired person can complete on a phone. The split also protects good work from looking useless: if voluntary churn falls while a processor issue pushes involuntary churn up, the blended rate stays flat and a successful retention project gets cancelled for failing at a job it never had.
Calculate the headline rate, absolutely. Then split it before anyone is allowed to act on it. A churn number should start a useful argument about causes — not hide two different causes inside one tidy percentage.
Published by Swipe Again — failed-payment recovery for subscription businesses.
Top comments (0)