Development of a Decision Support System using Data Analytics ...

We present a new decision support system for predicting customer churn for a leading online fast moving consumer goods retailer. The proposed system is based on recent data analytics and machine learning tools. The main steps of our development are data cleaning, feature extraction and prediction with supervised learning methods. The prediction step requires a careful definition of a churning customer in the retail sector for it is not obvious to identify the agents in such a non-contractual setting. After discussing different scenarios for customer attrition, we conduct an empirical study by using the past data of the company. Our results and observations contribute to the existing literature along two dimensions. First, we present the proposed decision support system as a promising application of data analytics to predict churn in online retail. As a matter of fact, the retailer in our study is currently in the process of implementing the proposed system within their existing infrastructure. Secondly, we answer two questions from the literature on the effect of the length of customer event history and staying power of prediction models. We argue that our results for the retail industry here validate the findings obtained for other industries studied in the literature.

Biography:

Ilker Birbil received his PhD degree from North Carolina State University, Raleigh, USA. He then worked for two years as a postdoctoral research fellow in Erasmus Research Institute of Management, Rotterdam, The Netherlands. His research interests lie in the intersection of large-scale optimisation and data analysis with a particular emphasis on algorithm development. He is one of the founders of the leading data analytics professional program in Turkey. He also serves as a board member in Center of Excellence in Data Analytics, Sabanci University, Istanbul. Currently, he is working as a faculty member in Sabanci University, where he teaches various courses on operations research.