A heuristic approach to in-season capacity allocation in a multi-product newsvendor model

Abstract Based on an actual problem, we develop and test two heuristics for optimizing the newsvendor’s response to new in-season demand information for a multi-product single-period problem. In this model, in addition to the quantities produced prior to the beginning of the selling season, the newsvendor can produce additional quantities of some products after the selling season begins. Sales data of all products is updated daily, and using these updates, the newsvendor can test if the quantities already produced are still optimal. If the newsvendor finds that the actual demand of a product up to a point in time indicates that the quantity already produced is too small, then the product becomes a candidate for an incremental production quantity. In Heuristic I, we fix the pre-season production quantities to those that are optimal for the classic newsvendor model. In Heuristic II, because the newsvendor may have some recourse during the selling season, we allow for reductions in the pre-season production quantities from the optimal classic newsvendor model production quantities. Using a simulation experiment, we show that even under an unbiased forecast, both heuristics result in higher profit, with Heuristic II outperforming Heuristic I. The increase in profits from using the heuristics increases significantly with the number of products for which the forecast is biased. In addition, the improvements in profits decrease significantly as the lead time and the penalty for late production increase.