Churn to the Power Three

Churn is probably the most critical concept to understand about your subscription business. In this article we describe:

  • 3 different types of Churn
  • 3 different ways to calculate Churn rate
  • 3 different ways to model Churn over time

Understanding all these different concepts and how they apply to your business is important. They represent different ways of thinking about Churn metrics and different gotchas in interpreting the results.

Three Reasons for Churn

  1. Inevitable Churn – Inevitable churn comes from customers who have already stopped using your product but not stopped paying for it yet. They no longer see/find value in what you provide and they will eventually cancel or their credit card will expire and they will never renew it.
  2. Avoidable Churn – This type of churn comes from customers who are on the brink of leaving. Often they will show signs of leaving such as reduced activity. This population is very interesting because these customers could be saved. If you can identify customers in this state you can call them and offer them support or a deal or training and even if they can’t be saved you can learn more about their issues that might help you save the next one.
  3. Seasonal Churn – seasonal churn occurs in industries where demand for the product  changes through the year. Seasonal churn is best addressed by having pricing options that allow for pausing the service at low cost during fallow times of year.

Three ways to Calculate Churn Rate

  1. Customer churn – the simplest and most obvious way to think if churn rate is to find the percentage of customers that leave in a given period (month/year). This is useful as a base metric but is not really representative of the loss to the business. If you have multiple plans with different costs the loss of a customer on a higher rate plan can have much more impact on revenue than the loss of several customers on the lowest rate plan.
  2. MRR Churn – a more sophisticated way to compute churn rate is to consider the loss of Monthly Recurring Revenue (MRR) and a percentage of total MRR. The benefit of this approach is that it normalizes for customers on different rate plans and gives a figure that represents short-term business impact. This is probably the most useful churn rate definition. One of the benefits of considering churn in this way is that cross sales and upsales can be combined into this single metric and if the increase in revenue due to these additional sales exceeds the loss of revenue due to customer departures we end up with a situation of negative churn. However, it fails to take into account the effect of churn over time on the overall customer value. For example the MRR calculation would value monthly customers higher than annual customers because their face value MRR is higher. However, we know that, over time, annual customers are usually much more valuable to the business (link).
  3. LCR Churn – churn can also be calculated using Lifetime Customer Revenue (LCR) [link] by considering lost LCR as a percentage of total LCR. This has the advantage of representing churn in terms of true business value. However , it does not provide the same level of immediate feedback on next months cash flow that MRR churn does. It is also more complex in the sense that it requires a model of churn over time to calculate LCR in the first place.

Three ways to model Churn over time 

  1. Churn rate – the simplest way to model churn over time is to come up with a single value for churn rate by averaging the loss of customers over the month. However, it is reasonable to suppose that the probability that a customer will leave varies over the lifetime of the customer for a given product. Customers may be much more likely to leave in the first few periods of use when they are still figuring a product out. Likewise some products may loose their usefulness or appeal after, say, a year.
  2. Churn by cohort – a more sophisticated way of modeling churn over time is to consider churn by cohort, we build a model of churn in which we consider the percentage of users who left after one month, two months, three months, and so on.  This kind of model gives a much more subtle model for churn that is particularly useful in predicting churn behavior in an environment where the acquisition rate varies significantly month to month.
  3. Churn normalized by sign-up anniversary – the final aspect of churn to consider in modeling is the distribution of sing up anniversaries. Subscription businesses with lumpy sign up history will experience lumpy churn. That is if there were a lot of annual sign-up in a given month because of a promotion or event it is important to expect more churn in that month the following year simply because more renewals will occur on that date.

We have discussed three reasons for of churn, three ways to calculate churn rate and three ways of modeling churn over time. Churn is a critical metric for a subscription business. When a business chooses its Key Performance Indicators (KPIs) it needs to consider these different alternatives. There is no right answer here its just important to choose an answer and understand the benefits and consequences of that choice.