CALL CENTER PERFORMANCE EVALUATION Tariq Omari Dept. of Systems and Computer Engineering, Carleton University, Ottawa, ON, K1S 5B6, Canada tomari@sce.carleton.ca Hussein Al-Zubaidy Dept. of Systems and Computer Engineering, Carleton University, Ottawa, ON, K1S 5B6, Canada hussein@sce.carleton.ca Abstract In this paper, the effect of using a combination of multi- skill and specialized agents on the performance of a call center is studied. An OPNET simulator for the call center has been designed, implemented, and verified. The designed simulator has the flexibility that facilitates comparison of different scenarios. The scenarios are mainly oriented toward finding the performance enhancement that could be gained by using a combination of multi-skill and specialized agents. As the usual case in such problems, there must be an optimum combination that results in the best performance for a lower cost. The designed simulator provides a very powerful and scalable tool that could be used to find such an optimum, and could be easily modified to support larger call centers. Some selected scenarios have been tested and the results introduced and analyzed. The result of our research concludes that the economies of scale could be obtained by cross training only a minor fraction of agents. Keywords: Call center; multi-skill; queuing model. 1. Introduction A call center may be defined as a service unit where a group of agents handles a large volume of incoming telephone calls for the purpose of sales, services, or other specialized transactions. Typically, a call center consists of telephone trunk lines, a switching machine known as the automatic call distributor (ACD), a voice recording unit (VRU), and telephone sales agents. Customers usually dial a special number provided by the call center. If a trunk line is free the customer seizes it, otherwise the call is lost. A fraction of calls that do not receive service become retrials that attempt to reenter service. Once the trunk line is seized, the caller is instructed to choose among several options provided by the call center via the VRU. After completing the instructions at the VRU, the call is routed to an available agent. If all agents are busy, the call is queued at the ACD until one agent is free. Once the trunk line is seized and until the caller leaves the system, any other customer cannot use the seized trunk line. Besides, an agent can service one caller at a time. Moreover, a caller remains in the system until it gets the requested service from an agent. This work determines the behavior of a call center when an increase in size and in customer population occurs, which is the natural evolution for any real call center. The available choice is to use only specialized agents. With only specialized agents we cannot profit from the economies of scale that arise when we have only cross-trained agents [2]. On the other hand, a growth can also be translated into an increase in the number of services they can provide. In such case, having multiple skills is the rule, not the exception. The use of multi-skill agents provides the solution but on the expense of extra training and wages costs. In this work, we try to find a mid point between these two extremes by using a mixture of these two to provide the best that could be gained of their benefits. Specialized agents cost less in the sense of wages, training requirements, management becomes easier in certain aspects, and they provide scalability for the call center. On the other hand, multi-skill agents cost more, need more training, and are less efficient in each individual skill, but they provide more flexibility in dealing with different types of services required. 2. Description of the Call Center The call center has three types of agents. The agents in each type have the ability to provide one, two, or three types of services. The call center is assumed to provide Banking (B), Insurance (I), and Travel (T) services. Each one of the three services has its own specialized software and database. The agents who provide more than one service need extra time to switch from one database to the other. The database switching time is assumed exactly 30 seconds. The calls arrive to the center with exponentially distributed inter-arrival time with mean 1/ . There are k trunk lines available. The VRU service time is assumed to be exponentially distributed with mean 1/µ. 3. Performance Measures There are two performance measures to be evaluated: 1. Quality of service: determined by the probability of blocking a customer (Pb), because of unavailability of trunks, and the average waiting time of a customer after dealing with the VRU till he talks to an agent. A 'good' quality of service is defined as having a low blocking probability, and a small average waiting time. 2. Efficiency: measured by the agent’s utilization and total cost of all agents working in the call center based on their salaries. The salary of an agent who provides only one 0-7803-8886-0/05/$20.00 ©2005 IEEE CCECE/CCGEI, Saskatoon, May 2005 1805