How PayPal fixed their 100 Million Dollar churn problem

As a product manager, prioritising churn management is essential. Churn, the loss of customers, can subtly yet significantly impact a startup. Timely identification and intervention are crucial, especially for startups where each customer loss is impactful. Recognising that some churn is natural and can reflect business growth, it's vital to address problematic churn effectively.

The Beginning: A Simple Approach to a Complex Problem

At PayPal, a significant churn problem was looming, costing an estimated $100 million. The solution came not from advanced technologies but through basic tools like a spreadsheet, SQL, and the analytical skills of a physicist named Ben Ramsden, whose problem-solving prowess was rooted in his physics background at Cambridge University.

Identifying the Real Issue

Initially, PayPal faced the daunting task of analysing why approximately 1,000,000 merchants were leaving each year. The first step was to distinguish genuine churn from misleading data. For instance, accounts that had stopped transacting but not officially closed were considered 'dark' rather than churned, as closing accounts were rare due to the absence of monthly fees. This observation highlighted the importance of recognising 'behavioural churn' - when customers gradually disengage before officially leaving.

Sifting Through the Data

The analysis then moved to exclude certain types of accounts like 'one-and-dones' - merchants who used PayPal briefly and then stopped. These were typically small, occasional users, not significantly affecting revenue. Another critical distinction was made between churn and onboarding issues, particularly with new signups having a bad initial experience.

Narrowing Down the Focus

PayPal also acknowledged 'false positives,' such as seasonal businesses, and 'non-regretted churn' involving users removed for policy violations. The focus eventually narrowed to established, consistent, non-seasonal merchants who suddenly stopped transacting. This refined focus revealed that the most significant revenue impact came from a small percentage of merchants.

Uncovering the Root Causes

The final stage involved deep dives into individual merchant histories to identify common issues. This task was undertaken by a diligent intern who sifted through various internal systems. About 20 critical scenarios were identified, leading to the creation of a simple yet effective predictive model.

Effective Solutions and Delighted Customers

The results of this analysis were regularly sent to customer service to proactively address emerging issues with merchants. This approach not only saved significant revenue but also greatly enhanced customer satisfaction.

Note: This story was shared by Matt on X via his thread