Understanding Your Customer Base: Churn

Ämne: Avancerad Analys

Churn is a means to understanding change in a business’s customer base due to loss of customers. Since keeping customers is generally cheaper than gaining new customers, then preventing the loss of customers is a profitable endeavor. The purpose of measuring churn, and similar metrics, is to evaluate actions taken by the business to retain customers or to identify weak points in the business.
Churn rate over time

Churn rate is a measure of churn in a given time period. Churn rate can be defined in multiple ways; here we’ll define churn rate as the ratio of customers lost in a given time period compared to the number of customers that could have been lost.

We can look at the extremes to understand churn rate better,

  1. If the churn rate is 0, then we have not lost any customers in the time period
  2. If the churn rate is 1, then we have lost all our customers in the time period

Below we can see the population at risk for each time period as green smiley faces, with red X’s over the population lost, i.e. the churned customers. On the lower axis, we display the calculated churn rate for each time period, \[\frac{\text{customers lost}}{\text{customers at risk to be lost}}.\]
Churn: Customers at Risk

Basic consideration for tracking churn rate:

    1. Define the moment a person becomes a customer (or user)
    2. Define when a customer has churned and is no longer a customer
      • deciding when a customer has churned can be subtle in some environments, such as eCommerce
    3. Determine the appropriate time period
      • the smallest gradient of time to track flux of customers; if you choose weeks it’s possible to aggregate up from weeks to months or years, but not down from weeks to days.
A Simple Example

Consider a subscription based business, where the customer pays in advance of receiving the service. Online services, phone companies, gyms and many other companies work in this way.

      1. A person becomes a customer on the date the service agreement is signed, and the first payment is received.
      2. The lost of a customer is set at the next billing date after the customer cancels the subscription, or the immediately after no payment is received.
      3. Since the billing cycle is monthly, a monthly time frame makes sense to record churn. A monthly period would imply there is a set billing date that is the same for all customers and likely a prorating of the first payment.
A More Subtle Example

A Swedish grocery store is an example of a commerce based business. It’s easy to understand grocery stores have repeat customers; it may be you buy all your groceries as the same store.

      1. A person becomes a customer when they sign up for a loyalty card.
      2. The customer is lost when there is zero spending on the loyalty card for a set amount of time. This amount of time can only be set by knowing and understanding your customer base, and your plan of action for preventing churn. A typical time period is 90 days.
      3. The time period for recording churn in this case is dependent on the set amount of time for zero spending. If 90 days is the limit for zero spending observing churn monthly is reasonable, however it may be advisable to consider shorter time periods to understand the effect of churn prevention programs.
Summary

Churn is flux in a customer base due to loss. Churn rate is a measure of the amount of churn in a predetermined time period. Calculating and recording churn rate is a fundamental steps in measuring, understanding, and reducing churn. Churn rate in this format is lets us look back and evaluate past actions, which leads to the question how do we look forward?

And Beyond

Looking forward with churn rate, we can calculate the expected churn rate for time period. Once we have the expected value of churn we can determine when churn is too high, or churn is reducing. One simple way to approach this is to use at averages and t-test, but a more accurate way is to apply some basic Bayesian analysis.

Going beyond churn rate, we can start look at calculating, measuring, and leveraging insight from tenure, hazard, and survival analysis. In future posts, we’ll review the fundamentals and some basic insights that can be gained form tenure, hazard and survival analysis. These three will give us insights on how and when churn is happening in the customers life cycle.