Profit-Based Model Selection for Customer Retention Using Individual Customer Lifetime Values.

Area of Science:

• Business Analytics
• Marketing Science

Background:

• Customer retention is crucial in competitive markets like telecommunications and banking.
• Existing customer churn prediction models often use a simplified Expected Maximum Profit (EMP) measure.
• The standard EMP measure assumes a uniform Customer Lifetime Value (CLV), which is unrealistic.

Purpose of the Study:

• To extend the EMP measure to account for individual customer lifetime value (CLV) heterogeneity.
• To provide a more accurate performance evaluation for customer churn prediction models.
• To offer improved decision-making support for customer retention strategies.

Main Methods:

• Developing an extended EMP measure that incorporates variable CLVs.
• Demonstrating methods to integrate CLV heterogeneity when CLVs are known, their distribution is known, or unknown.
• Applying the enhanced measure to assess the performance of churn prediction models.

Main Results:

• The proposed approach provides novel insights into model performance by considering individual CLVs.
• The enhanced EMP measure offers a more realistic evaluation of retention campaign profitability.
• The method adapts to varying levels of information about customer CLVs.

Conclusions:

• Incorporating individual CLVs into the EMP measure significantly improves the evaluation of customer retention campaigns.
• This data-driven approach aligns with modern business analytics and supports informed decision-making.
• The enhanced methodology offers a more accurate and actionable framework for businesses aiming to reduce customer churn.