impact of inventory centralization/decentralization on safety stock for two-echelon systems, The

Journal of Business Logistics, 1996 by Caron, Franco, Marchet, Gino

The allocation of safety stock inside a two-echelon distribution system made up of a central warehouse plus several remote ones (see Figure 1), involves both designing of a new system and reallocating of stock inside an existing one. Rosenfield and Pendrock1 pointed out that safety stock can be allocated in a distribution system following two alternative approaches:

1. In an independent system, the central warehouse safety stock protects the whole system against demand variations during procurement (or production) lead time (P), whereas the remote warehouse safety stock provides protection during transit lead time (T) only.

2. In a coupled system, the remote warehouse safety stock provides protection against demand variations during both procurement and transit lead times (P T), whereas the central warehouse acts as a staging area and does not carry any safety stock.

In the case of order-point systems, which are widely used in industrial inventory

management, procurement lead times are equal to the order-delivery cycle time between the central warehouse and the supplier, whereas transit lead times are equal to the order-delivery cycle time between remote warehouses and the central warehouse. The choice between an independent system and a coupled system has a great impact on safety stock holding costs, which are a main component of overall distribution costs. As to the amount of safety stock, the difference between the two systems can range from 10% to 50% for a distribution system made up of about 20 remote warehouses, given a low level of correlation between local demands (see Figures 2a and 2b).

An independent system is characterized by a location-based demand aggregation leading to savings in safety stock, since the central warehouse is subject to the overall demand variations of all geographic areas. Conversely, a coupled system (where safety stock is located at remote warehouses only) is characterized by a time-based demand aggregation, since its stock is capable of meeting local demand variations during the whole time span resulting from the addition of procurement and transit lead times.

The major factors influencing a choice between these two alternatives can be summarized as follows:

Number of remote warehouses. Since the higher their number, the more the benefits will result from the location-based aggregation, which is a typical feature of independent systems (i.e., demand fluctuations in a high number of geographic areas can more easily balance each other);

Procurement/transit lead time ratio. The lower such ratio, the more suitable is using a coupled system, which can benefit from the safety stock reduction ensuing from its time-based aggregation (i.e., the longer the transit lead time compared to the procurement lead time, the more easily demand fluctuations in different time intervals can balance each other);

Geographic correlation between local demands at different remote warehouses. Since in the case of correlated demands, a location-based aggregation leads to negligible safety stock savings (i.e., demand fluctuations in different geographic areas are not able to balance each other), thus making a coupled system more desirable;

Time correlation of local demands in each individual remote warehouse. This situation is similar to the previous one, since time-correlation decreases the benefits ensuing from a time-based aggregation (i.e., demand fluctuations in succeeding time intervals are not able to balance each other at a given warehouse), thus making an independent system more suitable;

Value added to a product during its transit from the central warehouse to a remote warehouse. The higher this value, the more desirable the storing of safety stock at the central warehouse, according to an independent system. This conclusion follows the postponement principle, which maintains that value should be added to a product as near as possible to the moment of its delivery to the final customer

Costs required by the transfer of goods between remote warehouses in order to make up for sudden variations of local demand. If these costs are much higher than those required to move goods from the central warehouse to a remote one, an independent system becomes the best solution;

Proportion of demand directly served by the central warehouse. The higher this proportion, the more desirable an independent system becomes, since it allows to keep a safety stock at the central warehouse.

Zinn, Levy, and Bowersox2 introduced the concept of portfolio effect in the case of two warehouses and analyzed the effects of inventory centralization/ decentralization on aggregate safety stock. Tallon3 extended the portfolio effect model to the case where supply or replenishment lead times are uncertain or variable. The purpose of this paper is to analyze the level of aggregate safety stock needed for an independent system and a coupled system in the presence of N remote warehouses by considering the effects of procurement and transit lead times and the correlation of local demands.