A Simple Six-Step Approach to Define Customer Churn in Retail

The retail industry is constantly in search of new ways to enhance the shopping experience of customers. In this article, let us spotlight customer churn and how data can be used to mitigate it.

What is customer churn?

When a customer leaves or stops transacting with the business, the business loses the opportunity for potential sales or cross selling. When a customer leaves the business without any form of advice, the company may find it hard to respond and take corrective action. Ideally companies should be proactive and identify potential churners prior to them leaving.

Why customer churn is important?

Customer retention has been noted as less costly than attracting new customers.

A key question in customer churn is, “Who are the active customers most likely to churn or decrease their basket size in the future?” One answer is to apply Advance Analytics to predict customer churn.

If a business can accurately identify the customers who have a high risk of churning, they can subsequently identify who out of these churners are high-value customers. Then targeted marketing campaigns can be run to get these customers to return to the store. By gauging the risk and value of the customers, a business can design and implement effective marketing campaigns to ‘re-capture’ these customers.

Why is it difficult to define churn in retail?

In the retail industry, the process of defining churn would be a difficult task relative to other industries such as telecommunications, insurance, etc. since it is a non-contractual business. As customers in retail have varied consumer patterns, the definition of churn would be different for each customer. There would also be the scenario of partial churn, as customers would not completely stop shopping but reduce their spend. An example would be a customer who might continue to buy dry goods from the supermarket but buy fresh items from a competitor.

The 6-step process to define customer churn in the retail sector

In this article, we will explain the process of defining the target variable (customer churn) before building the predictive model. One of the vital steps in predictive modeling, be it a classification or a regression approach, is the target definition.

With all this in mind, we could define customer churn in a six-step process.

There are two types of models to address the issue of partial churn — one for soft churn and one for hard churn. Therefore, two target variables columns will have to be created separately. Let us get started!

Step 1:

Customer Spend is aggregated at a weekly level.

Figure 1: Customer weekly spend

Step 2:

Calculate the moving average of the customer weekly spend over the last 12 weeks to smoothen out the fluctuation. You may use a different length of time to smoothen out the fluctuation, but in our scenario, 12 weeks would be the ideal period.

Figure 2: 12 Week moving average of the net sales

Step 3:

Calculate the standard deviation of the moving average of the past 24 weeks. Since 24 weeks is approximately 6 months of the customer purchasing cycle, this would be an adequate period to identify any fluctuations from the normal pattern. Going too far behind would mean not accounting for the natural changes in spending of the customer.

Figure 3: Standard deviation of the past 24 weeks of the moving average

Step 4:

Define a minimum threshold. We do this by subtracting the standard deviation from the moving average as per the following formula:

(12 week Moving Average - (24 week standard deviation * 2)

Figure 4: Lower bound of the moving average

Step 5:

Move the minimum threshold by 6 weeks to take pro-active action even before the customer churns. In the retail industry, the customer would not churn immediately. It would be a case of partially reducing the spend before finally moving the whole basket to a competitor store. Identifying the point in which the customer starts to churn beforehand leaves the marketing team with enough time to proactively implement interventions.

Figure 5: Lower bound lagged for 6 weeks

Step 6:

When the moving average goes below the lagged minimum threshold, the customer is identified as a churner.

Figure 6: Churner is identified when the moving average goes below the lagged lower bound

Great! We have managed to define churn for each unique customer.

As I mentioned earlier in this article, there are two types of churners — soft and hard. Below are the two ways in which these churners are differentiated.

· Soft churn is when the moving average goes below the minimum threshold defined by us.

· Hard churn is when the 12-week moving average reaches zero. This implies that the customer has not visited the store at all.

We need to be careful of customers who exhibit natural changes in their spending. How can we take this into account? If a customer is a soft churner for a period of approximately 3 months, we assume that maybe he/she has changed his/her behaviour and may not be a churner. In these cases, we could set a rule that if he/she is a churner for 3 consecutive months, he/she is no longer a churner.

Hopefully, this article gives you an insight into identifying customer churn in the retail industry, which assists in designing the right type of marketing intervention!

Written by Asanga Gunawardena, Data Scientist