Effect of granularity of resource availability on the accuracy of due date assignment

Abstract

In a make-to-order environment, order due dates can be assigned by consecutively estimating for each task on the order the time that the required resource will be available to perform the task. This paper examines the nature of the trade-off between the granularity of representation of resource availability, the resulting accuracy of due dates, and the computational time required to compute a due date. Granularity is defined as the size of the intervals of time (time buckets) over which available capacity is computed in a discrete-time representation of resource availability. The literature does not provide guidance on setting this important parameter, and typically it is arbitrarily set to one day or one week. Results compiled from extensive computations show consistent patterns of behaviour across all six combinations of values of experimental parameters for order size variability and system utilization. The resulting behaviour is classified into seven regions and analysed. Interestingly, results indicate that as granularity decreases the accuracy increases only up to a point, and the accuracy for very small granularities actually can be much worse than the accuracy of moderate granularities. From the results it can be concluded that when granularity is set to an appropriate value, resource availability will be estimated with relatively high accuracy and relatively low computational time, which are important performance characteristics for an order promising application.