Accuracy of noncomplex customer lifetime value models in the medical service context

Abstract

Objectives:

The accurate valuation of a customer relationship remains a challenge that researchers and companies alike are struggling to solve. The objective of this study was to assess the accuracy of noncomplex and deterministic customer lifetime value (CLV) assessment models in a semi-contractual service business context.

Methodology:

The data set encompassed four years of longitudinal behavioral data from 150 customers of the case company. Six noncomplex CLV models were selected and used to (1) predict the CLV's of individual customers, (2) to sort the customers into four equally sized segments based on the rank order of their predicted CLV, and (3) to predict the combined CLV of the customer base. The predictive performance of the models was evaluated by comparing the predicted CLV's with the actual values calculated from a holdout sample.

Findings:

Four models were found to be quite inaccurate and the remaining two models very inaccurate at predicting the CLV of individual customers. Four models were found to be somewhat accurate in sorting the customer base into four segments and more accurate in predicting the top 25% of customers. One model was also especially accurate in predicting the combined value of the customers and can thus be utilized in the business context of the case company for customer base valuation purposes.