A Systematic Approach to Segmentation Analysis Using Machine Learning for Donation-Based Crowdfunding

Abstract

Donation-based crowdfunding rose in popularity in the last five years. Yet the lack of customer provided information crowdfunding platforms challenged segmentation analysis for marketing decision-making. Our study solved a common problem that many donation-based crowdfunding platforms faced in segmentation analysis in the absence of donor descriptive attributes as well as unobserved customer heterogeneity. Using data from the world’s only donation-based crowdfunding charity, we segment online donors’ behavior with machine learning clustering algorithms by combining recency, frequency and monetary values (RFM) metric with donors’ campaign preferences. We identify and compare the optimal number of donor clusters from different clustering algorithms. This segmentation analysis with machine learning, we termed ‘RFMP’ framework, offers a robust and novel approach to parsimoniously segment online behavior using observed and unobserved customer heterogeneity when extant segmentation strategies typically apply RFM metric in combination with demographic and socio-economic attributes to infer and predict customer behavior. The resulting donor clusters inform marketing decision-making in the design and implementation of crowdfunding campaigns for charitable causes. Our project contributes to the application of machine learning in marketing strategy.