MANUFACTURING & SERVICE OPERATIONS MANAGEMENT Vol. 14, No. 3, Summer 2012, pp. 386–401 ISSN 1523-4614 (print) ó ISSN 1526-5498 (online) http://dx.doi.org/10.1287/msom.1120.0378 © 2012 INFORMS Robust Design and Control of Call Centers with Flexible Interactive Voice Response Systems Tolga Tezcan Simon Graduate School of Business, University of Rochester, Rochester, New York 14627, tolga.tezcan@simon.rochester.edu Banafsheh Behzad Industrial and Enterprise Systems Engineering, University of Illinois at Urbana–Champaign, Champaign, Illinois 61820, behzad1@illinois.edu W e consider the robust design and control of call center systems with flexible interactive voice response (IVR) systems. In a typical call center equipped with an IVR system, customers are handled first by the IVR system, which is a computerized system that often enables customers to self-serve. Those customers whose service cannot be handled by the IVR system are sent to the second stage to be served by agents. To address the design of a call center with a flexible IVR system, we consider a call center whose IVR system can operate in two different service delivery modes that have statistically different outcomes. We formulate a stochastic program to determine the number of agents needed in the second stage as well as the proportion of customers that should be selected for each IVR service mode under different demand levels with the objective of minimizing the total staffing and abandonment costs. We also propose dynamic control policies to achieve the optimal proportions without the knowledge of the exact arrival rate. We show that the staffing levels found by the stochastic program combined with the proposed routing policies are asymptotically optimal in large systems. We present extensive numerical and simulation results to illustrate the effectiveness of the proposed solutions. Key words : call center management; queueing theory; stochastic methods History : Received: March 21, 2011; accepted: November 22, 2011. Published online in Articles in Advance May 4, 2012. 1. Introduction Call centers have become one of the primary ways companies interact with their customers. Tradition- ally, the main service provider in a call center has been a call center agent. However, with the advent of new computer technologies, companies have adopted ways that customers can “self-serve.” One of the main technologies developed for this purpose in call centers is the interactive voice response (IVR) systems. These systems are also known as spoken dialogue systems, voice user interfaces, or speech applications. An IVR system is the front end of a call center that is a com- puterized system enabling customers to interact with the company without the assistance of human agents. Advanced IVR systems have the capability to auto- mate a significant portion of the service provided by call centers in different ways. Depending on the nature of the incoming calls, an IVR system can han- dle 20% to 60% of all calls (Dawson 2004), and they are also very cost effective in managing some of the routine operations such as call routing. Because the cost of handling a call (or a part of it) in an IVR system is a fraction of the cost of an agent handling the same call, most modern call centers use these systems. Successfully implemented IVR systems may lead to improved customer and call center agent sat- isfaction, increased revenue, and reduced cost; see Aspect Communications (2003). Some customers may even prefer not to deal with a human (Horovitz 2003). Despite the advantages of IVR systems, they are well known to be a major source of frustration for call center customers (see Aspect Communications 2003), influencing the customers’ overall perception of the company they are calling. There are now websites, e.g., http://www.gethuman.com and http://www .get2human.com, that are dedicated to providing ways callers can avoid the IVR systems of different companies’ call centers in order to opt out immedi- ately upon arrival and speak with an agent. Although, with the standardization of IVR software, designing and implementing IVR systems require minimal tech- nical knowledge (Cleveland 2006), there is a lack of operational understanding as to how IVR systems should be designed (Gans et al. 2003); see Paek and Pieraccini (2008) for a description of the design cycle of an IVR system. Customer frustration with IVR sys- tems has long been recognized, and there is a signif- icant research effort to improve the design and the technology of these systems; see Acomb et al. (2007) for a brief history of the technological improvements related to IVR systems. 386