Inventory Models Without Explicit Fixed Ordering Costs

Our research investigates new methods for solving single and multiple period, finite-horizon, inventory problems when the number of orders allowed during the horizon is limited. Existing literature has solved the question of how many goods to order when there is only one order available to make at the beginning of the horizon, as well as how many goods to order when there is an explicit and known fixed ordering cost incurred upon the placing of an order. Our models extend the problem to the case when the fixed ordering costs are either unknown or very difficult to estimate.;The classical "newsvendor problem" in the management sciences solves for the optimal ordering quantity when there is a single period with a known single demand distribution. Our first research question is how to extend this classical model when there exist two opportunities for ordering within the horizon. We develop analytical formulations to solve this extension optimally.;An extension of the single-period case with one replenishment is a general multiple-period model with multiple (restricted) replenishment opportunities, without an explicitly stated fixed ordering cost. To solve this problem, we employ a Markov Decision Process-based solution methodology, where the costs of over-ordering and under-ordering are weighed against each other to determine the optimal ordering quantity for any period, when the decision maker has a given number of orders remaining. We also investigate an approximation to solve this problem near-optimally. We extend this model to account for price markdown opportunities.;Our research also solves a related problem when the demand distribution parameters are unknown. We evaluate the extent to which additional orders can alleviate some of the risks involved in ordering under these assumptions. To solve this problem, we employ a joint simulation and Markov Decision Process solution methodology that utilizes both optimal and near-optimal solutions for the known demand case in a simulation environment consisting of unknown seasonal demand. We show the profit difference in the known and unknown demand parameter cases when multiple orders are available.