Allocation Planning for Demand Fulfillment in Make-to-Stock Industries - A Stochastic Linear Programming Approach

Demand fulfillment is a planning process which is concerned with the processing of customer orders. Its main objectives are providing a high customer service, especially in terms of providing real-time order confirmations and promising reliable delivery dates, as well as maximizing profits. In make-to-stock industries, such as the consumer goods industry, production quantities are usually determined mid-term, based on forecasts and not on actual customer requests. As a consequence, bottleneck situations can occur in the short-run. Furthermore, customers are usually heterogeneous regarding their profitability and their strategic importance. Consequently, the firm has to decide carefully on which orders to accept and whether to fulfill an accepted order from stock or from future production quantities, which entails either inventory holding or customer-specific backlogging costs. The setting in make-to-stock demand fulfillment is comparable to the situation in service industries: capacity is scarce in the short-run, customers are heterogeneous, and demand is uncertain. Therefore, quantity-based revenue management ideas have been transferred to the context of make-to-stock industries. However, typical revenue management assumptions like the perishability of products do not hold in the make-to-stock context. As a consequence, the allocation planning problem’s complexity increases. Existing allocation planning models for make-to-stock industries show two main drawbacks: they either do not consider information about demand uncertainty appropriately or they are not scalable and, thus, not applicable to problems of practical sizes. Commercial advanced planning systems also provide the opportunity of determining allocations by means of simple rules. However, they also do not consider information about demand uncertainty (nor about customer heterogeneity) appropriately. The dissertation shows how allocation planning for demand fulfillment in make-to-stock industries can be improved by means of two-stage stochastic linear programming (SLP) with recourse. SLP formulations for both the single-period and the multi-period case are given. Moreover, the dissertation illustrates that allocation planning can be further improved if information about the consumption process, which follows the allocation planning process, is integrated into the allocation planning SLP model. In particular, information about the order arrival sequence as well as about consumption policies such as nesting or time-based policies is integrated by means of the second stage. The benefit of both performing allocation planning and considering information about demand uncertainty by using two-stage SLP depends on the input data like, e.g. the degree of customer heterogeneity, of capacity shortage as well as of demand uncertainty (or the forecast accuracy, respectively). Within a numerical study, we evaluate when allocation planning is likely to be beneficial and, in case it is, when SLP models are likely to outperform the allocation planning rules of commercial advanced planning systems or other existing allocation planning models.