This paper develops a model for allocating cross-trained workers at the beginning of a shift in a multidepartment service environment. It assumes departments are trying to maximize objective functions that are concave with respect to the number of workers assigned. Worker capabilities are described by parameters that range from zero to one, with fractional values representing workers who are less than fully qualified. The nonlinear programming model presented is a variant of the generalized assignment problem. The model is used in a series of experiments to investigate the value of cross-utilization as a function of factors such as demand variability and levels of cross-training. Results show that the benefits of cross-utilization can be substantial, and in many cases a small degree of cross-training can capture most of the benefits. Beyond a certain amount additional cross-training adds little additional value, and the preferred amount depends heavily on the level of demand variability.manpower scheduling, service operations management, mathematical programming

This paper presents a new method, called the batch quantile method, for estimating quantiles in regenerative simulations. The quantile estimator is consistent and asymptotically normal, and the method can be easily implemented and does not require prior knowledge of the range of values the data will assume. Empirical studies show adequate coverage of confidence intervals when batches of 50 cycles or more are used.quantile estimation, discrete event simulation, regenerative stochastic processes

No abstract

For many financial models implemented in electronic spreadsheets, input data values frequently are random variables because they are actually estimates of unknown quantities. As a result, the bottom-line performance measure of the model is a random variable, and risk is associated with decisions based upon it due to the uncertainty in its value. We describe in detail how to evaluate this risk using simulation in a spreadsheet and illustrate the procedure with an example. Formulas for generating random variates from many common distributions using LOTUS 1-2-3 are given, and data analysis considerations are also discussed.

No abstract