Abstract
We study a new class of inventory control problems, the recruitment stocking problem (RSP), applicable to general recruitment systems and products with limited supplies. Recruitment stocking occurs routinely in many organizations with the goal of identifying qualified candidates rapidly and cost effectively. This activity may take place at multiple sites simultaneously to shorten the recruitment time. Examples include recruiting patients in clinical trials, enlisting personnel in the military, and recruiting customers for product sampling in market testing and promotion. RSP differs from the extant inventory management literature in that it stipulates a finite recruitment target, so that recruitment is terminated as soon as the total number of recruits across all locations reaches that prescribed target. This distinctive feature of RSP calls for the development of new stochastic models to evaluate and optimize system performance.
Thus, we present a novel methodology of relaxation and decomposition to characterize the probability distribution of rejections in RSP (number of arrivals to an empty inventory). This method provides a basis for efficient and accurate evaluation of the Type 2 service level and expected recruitment time. We also leverage the attendant computational efficiency to develop optimization algorithms to compute the optimal stocking quantities.