Implementing the Pareto/NBD model: A user-friendly approach

Abstract

The paper addresses an issue that has been qualified as challenging in the literature, i.e., estimating the parameters of the well-known Pareto/NBD model. Previous researchers have circumvented th e apparent difficulty by bringing slight variations to the model. These changes turn out to be less innocuous than initially thought. The study proposes an alternative method to compute the parameters of the Pareto/NBD model with maximum likelihood. The method relies on the evaluation of a single one-dimensional integral. It is illustrated with two data sets that have been used in the literature. The procedure leads to the same parameter estimates as the previously published ones. It can be implemented in a spreadsheet.