Funded Phd Research

Liangbin Katie Yang

Sequential Allocation for Customer Acquisition

Marketing Department; Faculty Adviser: Peter Fader

Customer acquisition has already been a critical process to every retailer but with today’s greater clarity about the value of customers (only after they are acquired). Over the years, retailers tend to focus primarily on the sheer quantity of customers and often use relatively naive acquisition heuristics simply because they don’t know how to predict the value of potential customers. Undoubtedly, retailers can bring much more efficiency to the key decision about how and where to constantly look for new valuable customers (before they are acquired). In particular, how should a retailer sequentially allocate its acquisition budget across segments to maximize customer profitability over time?

To answer this fundamental question, we examine a set of common acquisition policies – and some new ones – and evaluate their capabilities in identifying the “right” customers. Beyond the typical acquisition methods, which all tend to be “backwards looking” (i.e., defining the “goodness” of a segment based purely on observed outcomes), we introduce several “forward looking” onces, which utilize forecasts of behavior/profitability for each segment when making allocation decisions.

From methodological perspective, we develop a novel CLV model and link it with multi-armed bandit heuristics to address the challenges arisen in sequential allocation for customer acquisition. From managerial perspective, we are among the first to systematically examine the sequential allocation process for customer acquisition, propose a flexible and useful “three-step” framework, and provide retailers with several easy-to-follow guidelines and practical recommendations for various acquisition scenarios

Publications

Yang, Liangbin Katie, Eric M. Schwartz and Peter S. Fader (2015), “Sequential Allocation for Customer Acquisition,” Working Paper. https://www.dropbox.com/s/qk22wn08c7dici9/9_SACA_2015_03_19.pdf?dl=0