Abstract

This paper is being realized as part of our last year in master’s degree in Economics. It aims at studying the firm’s most valuable asset: its customers. To that end, we adopt a quantitative approach based on a mix of Econometrics and Data Science techniques with a threefold purpose:

  • Model customer portfolio as a set of customer segments.
  • Predict and analyse customer attrition.
  • Estimate the overall value of the firm’s portfolio.

After having defined the subject’s key concepts, we apply unsupervised classification algorithms and duration models to a kaggle dataset related to customers of a fictional telecommunications service provider (TSP). Once the customer segments identified and customer lifetime estimated, a new metric is proposed to compute customer value during their lifetime in the portfolio. The value of the portfolio is then calculated and various simulations are realised.

Keywords: customer portfolio management (CPM), churn, customer value, duration models, segmentation, telecom.