Abstract:

In the field of customer base analysis, numbers of probability models are established according to different behavioral characteristics that customers have showed in the transaction process. Such models characterize customer behavior by selecting different probability distributions in the model assumptions. Based on the five assumptions in the classic Pareto/NBD model, the differences and conversion conditions of various distributions in those assumptions are analyzed, both from the perspectives of mathematical properties and behavioral characteristics, and also the benchmark of selecting distributions in each assumption is summarized, so as to help modelers understand and distinguish the suitable scenarios for each distribution to characterize.

Key words: customer base analysis, probability distribution, customer behavior, customer lifetime value