Multi-Level Inventory Systems

Compared to single-level inventory systems, multi-level supply networks are much more complicated. The reason is:
In a multi-level supply network, an upstream node is the supplier of a downstream node. To put it the other way round, the downstream node is the customer of the upstream node. Therefore the replenishment lead time seen by the customer node is the customer order waiting time provided by the supplier node. Reducing the customer order waiting time is possible by increasing the supplier's safety stock. Reducing the customer's replenishment lead time reduces the required safety stock.
The majority of approaches to the analysis of supply networks consider system structures which include
solely inventory nodes. In the case studied most often, a two-stage inventory system is considered with one central warehouse (wholesaler, node located upstream) and multiple regional warehouses (retailer, nodes located downstream) serving random demands. The central warehouse sends replenishment orders to an external supplier who delivers after a fixed delivery lead time (which is the replenishment lead time of the central warehouse).
The inventory node located upstream (central warehouse) has several options to use the data that are available after an inventory review:

Local Control – Installation Stock Policies. An installation stock policy uses only
inventory data with respect to the local inventory node. In this case the replenishment
decisions are made for each inventory node in isolation, i. e. without considering
the states of the other nodes in the supply network.

Central Control – Echelon Stock Policies. Echelon stock policies use inventory
information from all nodes located downstream in the supply network which are
linked to the focused inventory node.

According to the different types of system structures which can be further differentiated with respect to the number of levels, we consider:

• Convergent systems
• Divergent systems
• Generally structured systems

In a convergent system each node of the supply network has at most one successor (downstream node). This type of system structure is often used for modeling the valueadding process of an end product in a supply network consisting of several tiers of suppliers and the different production stages within a factory.

In a divergent system each node has at most one predecessor node (upstream node) but may have multiple successor nodes. The most popular system of this kind is the ”One-Warehouse-N-Retailer” distribution system.

Finally, in a generally structured system there are no restrictions with respect to the number of predecessors and successors that a node may have. This system structure is the most complicated one and the structure that is most often found in practice, at least if the value-adding processes are modeled in detail.