What is churn rate?

Churn rate describes the departure of customers or revenue from a base that is assumed to be stable. The concept comes from subscription businesses, where every month there is a set of active accounts and part of that set stops being active. In English it is called churn; other labels in use are cancellation rate, attrition rate and customer loss rate.

It should be kept apart from similar metrics. Bounce rate counts sessions without interaction and belongs to traffic analysis, not to the commercial relationship. Retention rate is its arithmetic counterpart when calculated over the same period and the same base, so retaining 92 % is the same as churning 8 %. Staff turnover applies the same idea to employees and belongs in a different dashboard.

Two different things sit inside the metric itself. Voluntary churn happens when the customer decides to leave. Involuntary churn happens when a payment fails, a card expires or a renewal breaks because of an administrative error. Both add up into the same percentage and call for opposite responses: one is addressed through product and service, the other through payment recovery. In e-commerce without subscriptions the term is used by analogy, and there nobody has explicitly cancelled anything: the figure depends entirely on how many days of inactivity each company decides to call a loss.

The arithmetic is a division of what was lost by the starting base, multiplied by one hundred.

Cancelación de clientes (%) =
  bajas del periodo
  / clientes al inicio del periodo
  * 100

Cancelación bruta de ingresos (%) =
  (MRR perdido + MRR reducido)
  / MRR al inicio del periodo
  * 100

Cancelación neta de ingresos (%) =
  (MRR perdido + MRR reducido
   - MRR ampliado - MRR reactivado)
  / MRR al inicio del periodo
  * 100

De tasa mensual a tasa anual:
  1 - (1 - tasa mensual) ^ 12

The difficulty lies in the decisions surrounding that division. The first is the reference base. Counting customers treats a ten-euro account exactly like a ten-thousand-euro one, whereas counting recurring revenue weights every loss by its value. The second is how sign-ups acquired during the period are handled. If they enter the denominator, the base grows and the percentage drops; if they are excluded, the result describes only the behaviour of the book of business you started with. The third is the length of the period, because a monthly rate does not become an annual one by multiplying by twelve: it has to be compounded, since each month applies to whatever the previous month left behind.

The revenue variants build on that. The gross version adds full losses and plan downgrades. The net version also subtracts expansions and reactivations, so it can come out negative when the growth of the customers who stay exceeds the loss from those who leave. Billing platforms let you configure what goes into the numerator and how a period that is still open gets estimated, so the same figure changes with the settings chosen.

Hence the practical problem: two companies can apply the same formula and publish figures that cannot be compared. It is enough that one measures accounts and the other revenue, that one excludes the month's sign-ups while the other includes them, or that one counts the loss on the day of the cancellation request and the other on the day the paid contract ends.

The figure decides where the money goes. If the customer base loses a steady share every month, spending on acquisition first covers that hole and only then produces growth. When the rate climbs, the operational question is how much of that budget returns more if it goes into support, into onboarding new customers or into recovering failed payments.

It also feeds the customer value calculation. The average life of an account is derived from the churn rate, so an error in the percentage carries over in full to the value assigned to each new customer and, with it, to the maximum the company can afford to pay to acquire one. A definition changed halfway through the year makes two quarters incomparable without anyone noticing it in the report.

Segmented, the metric supports product and pricing decisions. If losses concentrate in an entry-level plan, in one specific acquisition channel or in the first three months of life, the conclusion differs from the case where they spread evenly across the base. In the first case there is a point that can be fixed; in the second, the problem usually sits in the fit between what was promised and what is delivered.

• Write the definition down before publishing the number: base (customers or revenue), length of the period, treatment of sign-ups within the period, and the exact moment an account counts as lost.
• Separate voluntary from involuntary churn and give each its own plan, because one is fixed through the product and the other through payment recovery.
• Measure by sign-up cohorts instead of looking only at the monthly aggregate, so you can see whether losses occur in the first weeks of life or much later.
• Always publish the percentage next to the absolute number of losses and the size of the base, because 10 % of 40 accounts and 10 % of 4,000 call for different decisions.
• Segment by plan, acquisition channel and account age before drawing conclusions about the whole.
• Review the definition whenever the pricing model or the billing cycle changes, and record the date of that change in the time series.

• Annualising by multiplying the monthly rate by twelve. Compounding gives a different result, and the gap widens as the rate rises.
• Quietly putting the sign-ups won during the period into the denominator, which automatically lowers the percentage as soon as the company grows fast.
• Comparing your own result against published percentages whose calculation method is unknown, and setting internal targets from that comparison.
• Reporting customer churn and calling it revenue churn, or switching between the two depending on which looks better that month.
• Stopping at the aggregate percentage and missing that nearly all the loss sits in one small segment.