Live-chat agent assignments in two-class e-customer queues under imperfect profiling

Many e-commerce firms provide live-chat capability on their websites to promote product sales and to offer customer support. With increasing traffic on e-commerce websites, providing such live-chat services require good allocation of service resources to serve the customers. When service resources are limited, firms could prefer to serve more valuable customers first in order to obtain the most revenue. In this paper, we consider a priority scheme to determine the order of customers to be served in an e-commerce live-chat system that employs imperfect customer profiling. Two decision variables considered in our priority model are: 1.) the number of servers exclusively reserved for preferred customers, and 2.) false positive and false negative rates selected for the profiling system. We find out that both relative expected values of the customer types and the profiler configuration play important roles to identify optimal policies. Given the imperfectness of the profiling system, high specificity configurations that lead to less falsely valued customers may not be the best strategy when reservation policies are already in place.