One of the important uses of an EOQ item's distribution of lead time demand is to set its reorder point (when to order new stock). The current Air Force system computes the reorder point by assuming that demand during lead time is normally distributed. The analysis presented here shows that a much more realistic model of observed demand patterns can be chosen from the compound Poisson family of distributions. The geometric-Poisson and constant-Poisson members of that family are used in this study. Also we allow for fixed and variable lead times. The ultimate objective throughout the study is to understand, as exactly as possible, the true underlying random processes involved in the EOQ supply system. A strong secondary objective, though, is to allow for a probability model that could be implemented in the environment of a large base supply account. Data from eight federal stock groups and four bases (three CONUS and one overseas) are analyzed. Approximately 10,000 items are involved. We analyze individual customer arrival and demand processes as well as empirical lead time data. The primary emphasis is on fitting the assumed geometric-Poisson and constant-Poisson models to the data. The geometric-Poisson model is taken to be a better representation of the actual arrival and demand processes than the constant-Poisson model. Although the latter model is a cruder approximation it does represent an easier model for possible implementation. Both models are shown to be more representative of the actual data than is the normal distribution assumption. Also we show that both models perform better in setting hypothetical reorder points than the current model. Only a cursory examination of the effect of variable lead tine is attempted since the data contain many outliers. The data are sufficient, though, for testing the sensitivity of the reorder point computation to variable lead time. Our analyses show the effect to be almost negligible for CONUS base û CONUS resupplier combinations. The effect for overseas base - CONUS resupplier combinations probably is significant. FORTRAN programs are available for all of the analyses described.

"April 1980." Also available as: ADA082997.