Africa

Identifying Pilgrim Bank’s Loyal Customer

Team 7:

Michael Isble Karl Olson Johan Rong Ampun Janpengpen Hyun Kim

Executive Summary

Our objective with this project was to see if we could tease out information from a dataset of customers at Pilgrim Bank, a small retail bank in Texas. The project was strictly pedagogical. We received the dataset from P.K. Kannan and had no access to Pilgrim’s management. The dataset contained nearly 32,000 data points and had eight variables – District: referring to one of three zip codes around the banks’ retail location, Profit: indicating how much the bank made from each customer in 1999, Age: broken down into seven categories, Income: broken down into nine categories, Tenure: a numerical variable indicating years at the bank as of 1999, Online: a binary variable indicating whether or not a Pilgrim customers banked online, Bill Pay: a binary variable indicating whether or not a customer used Pilgrim’s online bill pay service, and Retained: our binary response variable indicating whether or not a customer was retained by Pilgrim in 2000. Approximately 20% of the data points did not list Age and Income. whole. While there is a strong correlation between

not listing this information and a customer not being retained, little else is can be deduced about this group as a None of the predictive variables is well correlated with the “not listers”.

Our analysis indicates that the district in which a customer banks has little bearing on whether or not they will be retained. This indicates that all three branches likely provide relatively similar service. Additionally, the use of Pilgrim’s online bill payment service did not seem to affect customer retention levels. On the other hand, customers who banked online were more likely to stay with Pilgrim. might be expected, tenure with the bank had an impact on retention. with Pilgrim, the more likely they are to stay with the bank. than are younger patrons. customers’ higher...