Filling customer orders from multiple locations: A comparison of pooling methods

Journal of Business Logistics, 1999 by Evers, Philip T

Demand and leadtime uncertainties are expensive. It is clear from the well-known formula for the standard deviation of leadtime demand that, when the safety factor (k(square)) approach to setting safety stocks is used, inventory increases as the standard deviation of either demand or leadtime increases. As a result, management may attempt to reduce this inherent variation by pooling it. Defined as advantages arising from uncertainty pooling, statistical economies of scale can be achieved in numerous ways, including inventory centralization, order splitting, and emergency transshipments.

Inventory centralization physically consolidates stock at a limited number of locations (often, a single facility) from which all demand is satisfied. Order splitting occurs when a stock-keeping location operates independently of all other facilities in filling its demand but divides (not necessarily evenly) its reorders among multiple supply sources. Emergency transshipments are used to fulfill demand by shipping product from facilities with excess stock to facilities with no stock. Previous work has found that inventory centralization results in demand pooling,1and order splitting results in leadtime pooling.2 Emergency transshipments have been found to result in a combination of both.3

To measure the demand-pooling effects of inventory centralization on safety stocks, the portfolio effect model was developed,4 It has been shown to hold in the case of demand pooling stemming from nonemergency transshipments (demand is partitioned among locations without regard to inventory levels).5 Because of the strong correspondence between inventory centralization and nonemergency transshipments, the use of emergency transshipments has been compared with inventory consolidation and found to outperform centralization on the basis of inventory availability.6 As part of that analysis, Evers noted that emergency transshipments bear some resemblance to order splitting due to leadtime pooling and speculated that they may do better than split orders in terms of inventory availability.

This hypothesis and the effect of these approaches on total cost drivers are examined here. After a discussion of the three approaches and a review of the literature, the methods used to investigate their relationship and the hypothesis are explained. The results are presented next, and managerial and research implications are addressed at the conclusion.

ACHIEVING STATISTICAL ECONOMIES

Emergency transshipments are employed in order to maintain inventory presence in local markets. By keeping decentralized facilities open and stocked, a firm is able to fill immediately a large proportion of demand from inventory, thereby preserving high levels of customer service through minimal average order cycle times. When demand is unexpectedly high and cannot be filled from on-hand inventory, an emergency transshipment from another facility with sufficient stock also preserves high levels of customer service through maximal inventory availability. (Unless otherwise noted, further reference to emergency transshipments will simply be termed transshipments.) Indeed, management may be able to support high customer service levels through the use of transshipments while at the same time reducing overall on-hand inventory levels at the decentralized locations; this is the notion of demand pooling.

Transshipments have the effect of pooling leadtimes, too. Individual locations place reorders independently, so a late-arriving order at one location may be offset by an early-arriving order at another. Coupled with the use of transshipments, replenishments arriving at different times in different locations can still be used to fulfill demand throughout the system; consequently, inventory availability improves.

Transshipments, however, are not free. A premium is paid to expedite small shipments between locations. Furthermore, the decentralized locations need to interact with one another in order to determine the appropriate facility from which to request a transshipment. Perhaps most important, management at individual locations must be willing to provide stock to others upon request. As long as these drawbacks in total do not outweigh the advantages of transshipment, its use should be permitted.

Much research on transshipments has focused on the mathematical modeling of particular systems. Work along these lines includes that of Krishnan and Rao,7 Lee,8 Tagaras,9 Axsater,10 Dada,11 Sherbrooke,12 Tagaras and Cohen,13 and Xu, Evers, and Fu.14 In general, it confirms that transshipments reduce stockouts by improving inventory availability. Simulation analysis also has been used to examine various aspects of transshipments (for example, Cantagalli, 15 Evers,16 and Needham and Evers 17).

A second method for achieving statistical economies, inventory centralization, lacks a key advantage of transshipments. By definition, consolidation leads to the withdrawal of inventory from markets and a fundamental restructuring in the logistics network of the firm. Inventory is held at a handful of locations, leading to longer order cycle times and/or higher transportation costs for distribution to markets no longer stocked locally. This technique typically requires considerable longrange planning because facilities are closed (and, possibly, new ones opened), inventories moved, and personnel displaced. Because centralization cannot usually be quickly implemented, it is not considered further in this study. To be sure, demand pooling occurs, and as a consequence, inventory availability rises while inventory levels fall. Consolidation alone, however, does not result in leadtime pooling and has been shown to be eclipsed by transshipments,ls which is a second reason for not exploring centralization further in this analysis (for more on consolidation, see, for example, Das and Tyagi 19).