How to Analyze Customer Churn Using Transaction Data

In a subscription business, churn is a cancellation. In stores, wholesale and services, customers rarely cancel; they just stop ordering. There, churn has to be defined with a rule, such as "no purchase within a set number of days".

Inactivity is the measurable fact: no purchase for a period. Churn is the conclusion that the customer has left. Some inactive customers return, so treat churn figures from transaction data as observed inactivity rather than certain loss. Revenow, for example, labels customers inactive after 90 days or more relative to the analysis reference date and does not use predictive churn models.

• Churn rate: customers who became inactive in a period ÷ customers active at the start
• Repeat purchase rate: share of customers with more than one order
• Revenue churn: the historical revenue of customers who became inactive

Group customers by the month of their first order and track how many are still buying later:

A falling share across cohorts suggests newer customers are being retained less well.

Illustrative example with hypothetical figures.

• Counting recent first-time buyers as churned before they had time to reorder
• Mixing very different customer types in one rate
• Changing the inactivity rule between reports

Once you know where churn is concentrated, focus on at-risk customers before they lapse. See how to identify at-risk customers and customer churn analytics.