Effective management of inbound telephone call centers / Yongpin Zhou.
xii, 212 p. ; 29 cm.
- Local subjects:
- Penn dissertations -- Operations and information management
Operations and information management -- Penn dissertations.
Penn dissertations -- Managerial science and applied economics.
Managerial science and applied economics -- Penn dissertations.
- This dissertation treats three operations problems in inbound call centers which are inter-related but on different time levels.
We first treat the long-term problem of staffing a call center under random, non-stationary service requirements. Training leadtime is long and total employee capacity changes systematically, but randomly, over time, due to employee learning and turnover. We use an MDP approach to explicitly model employees at different speed/skill levels. We show that a "hire-up-to" policy is optimal for both discounted and average-cost models. We also develop structural properties of the optimal policy. For example, we show that myopic policies are optimal under certain conditions.
Numerical analysis results suggest that when the learning curve is not steep or when flexible service capacity is easy to obtain, simple heuristics based on myopic policies work well. In the other cases, a more sophisticated model such as ours is necessary. For these cases, one may use simulation-based optimization methods that utilize problem convexity to quickly converge to the optimal up-to levels.
We then study a special version of the short-term skills-based routing problem and how it is embedded in and influences workforce scheduling. For this structure, which is common in practice, we show that, "threshold-reservation" policies perform well---and sometimes optimally---in a variety of circumstances. Furthermore, we develop algorithms to determine control parameters and to derive system performance measures. These algorithms are efficient and quick, making them suitable for repeated use within the larger workforce scheduling problem.
We finally study a single-server queueing system with service times modulated by a 2-state Markov chain. A closed-form solution of the steady-state number-in-system distribution is given. It has a matrix geometric form that holds for both finite and infinite waiting space cases, and it extends the existing results in the literature by Neuts  and Naoumov . We then use this system to investigate human factors such as learning and turnover. Numerical results suggest that, depending on the learning and turnover rates, the human factors may have a positive or negative effect on the system performance. We present a conjecture that generalizes these results.
- Adviser: Noah F. Gans.
Thesis (Ph.D. in Operations and Information Management) -- University of Pennsylvania, 2000.
Includes bibliographical references.
- Local notes:
- University Microfilms order no.: 99-76497.
- Gans, Noah F., advisor.
University of Pennsylvania.
|Location||Notes||Your Loan Policy|
|Description||Status||Barcode||Your Loan Policy|