Vijay Mehrotra

A Call Center Uses Simulation to Drive Strategic Change

A large, customer-focused software company relied on simulation modeling of its call center operations in launching a new fee-based technical-support program. Prior to launching this rapid program, call center managers were concerned about the difficulty of meeting a proposed guarantee to paying customers that they would wait less than one minute on hold. Managers also wanted to know how the new program would affect the service provided to their existing base of regular, nonpaying customers. We quickly developed an animated simulation model that addressed these concerns and gave the managers a good understanding for the impact on system performance of changes in the number of customers purchasing the rapid program and in the number of agents. The one-minute guarantee would be fairly easy to achieve, even if the percentage of callers in the rapid program became quite high. Managers also gained confidence that, with appropriate staffing levels, they could successfully implement the new program, which they soon did.

Forecast errors in service systems

We investigate the presence and impact of forecast errors in the arrival rate of customers to a service system. Analysis of a large dataset shows that forecast errors can be large relative to the fluctuations naturally expected in a Poisson process. We show that ignoring forecast errors typically leads to overestimates of performance and that forecast errors of the magnitude seen in our dataset can have a practically significant impact on predictions of long-run performance. We also define short-run performance as the random percentage of calls received in a particular period that are answered in a timely fashion. We prove a central limit theorem that yields a normal-mixture approximation for its distribution for Markovian queues and we sketch an argument that shows that a normal-mixture approximation should be valid in great generality. Our results provide motivation for studying staffing strategies that are more flexible than the fixed-level staffing rules traditionally studied in the literature.

Performance measures for service systems with a random arrival rate

It is commonly assumed that the arrival process of customers to a service system is a nonhomogeneous Poisson process. Call center data often refute this assumption, and several authors have postulated a doubly-stochastic Poisson process for arrivals instead. We develop approximations for both the long-run fraction of calls answered quickly, and the distribution of the fraction of calls answered quickly within a short period. We also perform a computational study to evaluate the approximations and improve our understanding of such systems.

Routing to Manage Resolution and Waiting Time in Call Centers with Heterogeneous Servers

In many call centers, agents are trained to handle all arriving calls but exhibit very different performance for the same call type, where we define performance by both the average call handling time and the call resolution probability. In this paper, we explore strategies for determining which calls should be handled by which agents, where these assignments are dynamically determined based on the specific attributes of the agents and/or the current state of the system. We test several routing strategies using data obtained from a medium-sized financial service firms customer service call centers and present empirical performance results. These results allow us to characterize overall performance in terms of customer waiting time and overall resolution rate, identifying an efficient frontier of routing rules for this contact center.

Successful modeling of a semiconductor R&D facility

A chief limitation on process development cycle time is the time it takes to process wafers. Hewlett Packard Co. has shortened this time by using discrete-event simulation of its R&D fabrication facility to better understand its capacity limitations and to analyze how changes affect the cycle times of complete wafer processing. Results include Pareto charts of equipment according to the impacts that changes in their mean-time-between-failure (MTBF) and/or mean-time-to-repair (MTTR) parameters have on wafer cycle times. Results also include a Pareto chart of operator skills and their impact on cycle times; that is, one can know how much effect there is on cycle times when staffing is changed to satisfy needs for a specific operator skill. Similarly, it is possible to evaluate effects of shift schedules and end-of-shift wafer processing decisions. Furthermore, simulation reveals the relative benefits of applying different dispatching rules (such as first in, first out, shortest-processing time, least-work-in-next queue, etc.) and of different inventory-level control policies.<<ETX>>

A manager-friendly platform for simulation modeling and analysis of call center queueing systems

Call center operational performance is measured largely through queue times and customer abandonment rates, and thus managers have an acute need to understand how both management policies and stochastic factors affect these performance statistics. Simulation is an excellent vehicle for examining these relationships, but a lack of programming ability can be a barrier that prevents call center managers from making use of such models. To address this problem, we have developed a user-friendly Excel interface for a dynamic discrete event simulation model. The underlying model is a general queuing system for which analytical results are often unavailable, and the Excel interface enables managers to interactively specify a wide range of system parameters and analyze results, all without exposing them to the simulation models components. Based on input from call center operations managers, we have also been able to utilize this framework to ask, and answer, some important empirical questions.

OR Process Skills Transform an Out-of-Control Call Center into a Strategic Asset

A large consumer-software company was struggling to manage a seemingly unmanageable, high-cost technical-support call center. The company used “OR process skills” to transform the call center into a strategic asset. By focusing on executive priorities, personally observing business processes, engaging with frontline workers, and directly examining the sources of important data, we discovered the central problem amidst a contentious, disorganized situation. We used a pilot program to test simple analytical tools, such as Pareto charts and sampling, to bring actionable information to the right parts of the organization. Following the processes we developed, the company analyzed customer feedback to improve the product and customer self-support mechanisms, thereby reducing both current and future call volumes. By empowering client staff and leading process change across functional boundaries, the company reduced its call-center costs and achieved higher product quality. In addition, we demonstrate that OR process skills can be an essential element in sustaining long-term consultant-client relationships.