LRFMVD : a customer segmentation model

Abstract

Customer segmentation is a big part of the superstore industry. Traditionally, the RFM model has been used to segment customers to maximize profit. This work proposes a new customer segmentation named LRFMVD based on RFM and LRFMV models in hopes of providing a more sure-fire way of segmenting customers. The k-means clustering method will be used for the proposed model. The clusters created by K-means are then analyzed using the LRFMVD model to find a correlation between profit and volume. Many works have been done previously on customer segmentation for maximizing profit, but none of those were able to show a straightforward representation of profit, volume, and discounts on products. Unsupervised learning was used to investigate the correlations between volume, discount, and profit. Customers are then segmented using the Customer Classification Matrix, which looks at the properties of all clusters. The L, R, F, M, VD parameters’ values are compared to the cluster mean values, and based on whether these values are higher or lower than the average, customers are segmented. Comparisons among the three models reveal that the latter provides more profit per head than the other two, and is able to identify customers who cause superstores to lose money or make a loss.