Matching Customer Inquiry with Agent Specialty: A Text-Analytic Framework to Reduce Routing Transfer Rates in Online Service Centers

Customer – service agent mismatch is a common problem in many contact centers leading to service rework (i.e., customer transfers), operational waste, and customer dissatisfaction that collectively cost firms millions of dollars each year. This problem looms larger as companies move their customer service channels online. Existing contact routing systems mainly rely on customers’ self-selection of problem category from a menu-based system. Consequently, unique characteristics of online communication such as free-form problem description texts are often under-exploited for customer-agent matching purposes. We propose a text-analytic framework that leverages problem description texts to improve customer routing accuracy and hence reduce transfer rates in online service centers. Grounded in the design science approach and built on computational linguistics and machine learning methods, the framework helps extract signal cues from text that can be used as modeling inputs to identify the true nature of a customer’s problem. To demonstrate the usefulness of the framework, we conduct a comprehensive case study on a large dataset collected from the online service center of a S&P 500 company. Our results indicate a 19% accuracy improvement from the framework over pure menu based routing. To assess the broader managerial implications of this improvement, we estimate potential reductions on agent service time and customer waiting time, as well as potential increments in customer’s willingness to recommend the company’s product. We find that the firm can save over $950,000 in labor costs and reduce the total waiting time of customer population by 9,600 hours per year, while significantly increasing customer satisfaction by switching to the new routing system.