Predicting Customer Churn using the Cumulative Quantity Control Chart

This study proposes a customized prediction scheme for customer churn. This scheme is based on cumulative quantity control (CQC) chart that monitors customers' inter arrival time (IAT). In addition, recency, a time interval pattern that is complementary with IAT is integrated in to increase false positive rate (FP) and to reduce false negative rate (FN) and average time to signal (ATS). Unlike the previous studies that are presented with static data analysis and tabular reports, CQC offers a unique prediction scheme that, in addition to graphic visualization, can perform dynamic monitoring as time passes and new information is collected. When a customer exceeds the control limit at a CQC score, the scheme issues an out-of-control warning for the bad behavior to help the administrator to take preventive measures. This paper conducts empirical analysis of the database of an online dating website in Taiwan and compares different CQC-v of Xie et al. (2002) with CQC of Chan et al. (2000), and the results show that the accuracy (ACC) of CQC-4 is the highest and ATS places second on the list.