Modeling and Optimization of Complex Service Delivery Systems

Introduction

The services sector represents a growing fraction of the U.S. and world economy. In fact, recent data from the U.S. Bureau of Labor Statistics reveals that services represents more than 70% of the total industry output (as measured in U.S. dollars) and more than 75% of the U.S. labor force is employed in the services sector. While employment in manufacturing has declined steadily since 1990, the number of employees and jobs in the services sector has grown and is anticipated to continue to increase in the upcoming years. This growth in the services sector has resulted in significant challenges to the viability and profitability of incumbent service providers, due to the growing number of competitors in the market. The new entrants challenge incumbent providers, often providing richer and more diverse offerings at lower cost. Customers have thus come to expect price reductions from service providers at the same high quality of service. Service providers must therefore seek ways to improve efficiency and reduce the cost of delivery while maintaining or increasing the quality of service. The cost of human capital is the most significant cost faced by a service provider. Thus it is critical for service organizations to invest in methods that support accurate determination of the number of resources with different skills that are required at different times and in different locations from which customers are serviced.

Resource requirement planning is a well-studied problem in the Operations Research literature. However, this problem has primarily been studied in the contexts where the resource being managed is parts or objects whose attributes do not change dramatically over the planning horizon. Even perishable items, whose attributes or whose capacity to service demand change over the planning horizon, differ from resource planning with human capacity. For example, people learn and acquire new skills, lose skills that are not used, and have service rates that may be impacted by the composition of their workload or by the sequence in which they process different types of requests. Thus, new methods are being developed to study and optimize service systems, beyond the methods that have been developed for traditional manufacturing settings.

The methods that we describe falls in the area of optimal staffing with skills based routing (“SBR”), where customer requests arrive with specific skill requirements and are subsequently serviced by agents with corresponding skills. The SBR problem is known to be analytically complex with limited theoretical results. Gans et al. (2003) and Aksin et al. (2007) provide detailed surveys of the analytical approaches (in the form of analytical queueing methods) that have been undertaken. The most common approaches are to either simplify the topology of the delivery network or simplify the routing schemes. However, none of them are desirable in a service delivery environment where both the network and the routing schemes are complex and the service providers are seeking practical solutions but not conceptual guidance.