Churn

Churn is the English word for attrition or turnover. It refers to customers who cancel an ongoing subscription and stop paying. Expressed as a number: out of 1,000 customers at the start of the month, 950 remain by the end of the month. That means churn stands at 5 percent per month. Nearly every company with recurring payments watches this figure closely. This includes streaming services, mobile phone plans, gyms, and providers of software that is rented monthly.

The hole in the bucket

A subscription business works like a bucket with an inflow and a leak. New customers come in at the top, while cancellations cause customers to flow out at the bottom. If a company is growing, that alone says little. What matters is how much is simultaneously flowing out at the bottom.

The reason for the fuss around this number is mathematics. 5 percent churn per month sounds harmless. But over the course of a year, only about 54 percent of customers remain. At 2 percent monthly, it’s around 78 percent. This difference determines whether a company grows or stays stuck in place.

On top of that comes the cost of acquiring new customers. Advertising, discounts, and sales cost money, often more than a customer generates in the first few months. A customer only becomes profitable after a longer period of time. If they cancel before that, the advertising spend was a losing proposition. That’s why investors often say churn is more important than growth.

How the number is calculated and predicted

The simplest calculation is customer churn: cancellations divided by the number of customers at the start of the period. There is also revenue churn. It doesn’t count heads, but lost money. This makes a difference when a single large customer pays as much as a thousand small ones. Some providers even report negative revenue churn. That means the remaining customers are spending so much more that it more than offsets the cancellations.

To predict cancellations, companies deploy machine learning models. Such programs learn from historical data which patterns typically preceded a cancellation. They are fed features such as usage duration, number of logins, support complaints, or plan changes. As a result, they produce a probability for each customer of canceling within the next month.

A common misconception is that such a model explains the causes. It does not. It only finds correlations in the data. The fact that a customer contacted support is an indicator, but not the reason for the cancellation. Anyone who draws the wrong conclusions from this ends up fighting symptoms instead of problems.

Churn in quarterly reports and in your own inbox

In business news, the term usually comes up around quarterly earnings. Streaming services report on lost subscribers, mobile carriers on contract cancellations. If churn comes in lower than expected, the stock price often rises. Among enterprise software providers, an annual attrition rate of under five percent is considered very good.

As a customer, you constantly encounter countermeasures without ever hearing the term. The discount offer shortly before your contract expires is one. The email with the subject line “We miss you” after weeks without a login is another. The many clicks required to reach the cancellation button also belong here, though this practice has since been restricted in the EU.

Churn should be distinguished from retention, i.e., the retention rate. Both describe the same thing from opposite directions. 95 percent retention corresponds to 5 percent churn. Companies like to cite retention because the larger number sounds friendlier.