Churn Analysis Modeling For Early Detection of Client Attrition And Pro-Active Marketing Efforts

Churn Analysis Modeling For Early Detection of Client Attrition And
Pro-Active Marketing Efforts

The Challenge

A large mobile telephony provider in Ireland had a problem. The cost of acquiring a customer was higher than the profit they made on a customer if the customer stayed with the company for less than 18 months. The company signed their customers into plans that tied the customer in for a period of 12 months but incentives offered by competitors and a new government directive that allowed portability of numbers meant that there was little reason for customers to stay longer than 12 months.

Hypothesis

  • Usage data can provide signals that help identify a potential churn.
  • Customers that churn due to issues with service provision show signs of disengagement well before the end of their tenure.
  • Usage data can also provide the basis for estimating the lifetime value of a customer.
  • Usage data can also provide insights that can be useful in personalizing offers to customers to “Save” them from churn.

Execution

  • Developed models of churn using survival analysis approaches.
  • Life Time Value was measured using the expected survival length and a regression model to predict expected profitability.
  • Features were extracted from behavioral data to enhance the models.
  • Unsupervised approaches were used to identify different usage patterns.

Outcomes

  • Models were deployed for the internal team to identify when a customer was at risk of churn.
  • Life Time Value estimates allowed the “Save team” to identify how much they could afford to spend to retain customers.
  • Visualization of different usage patterns helped the “Save team” to build offerings that were personalized to customer cohorts.

Project Highlights

30%

AVERAGE INCREASE IN LIFE-TIME VALUE