A Better Approach for Changing Your Supply Chain Network and Balancing Inventory Levels

Supply chain network design has two key functions: Determining the impact of manufacturing and distribution network changes on inventory levels and determining the storage space that inventory will require. That’s why most commercial network design software has functionality to examine this relationship. Knowing exactly what you’re optimizing is essential to getting the right answer.

You can only optimize one business problem at a time. Be sure to pick the right one

Supply chain modelers and fellow math enthusiasts may see the obvious “chicken and egg” paradox in combined network and inventory optimization. Mathematically, you can only optimize one business problem at a time. In the supply chain, you can design the best network configuration based on projected inventory levels. Or you can determine the right amount of inventory to hold, given your network parameters and desired service levels. Usually, you can’t do both, at least not at the same time.

In most network modeling applications, this math problem is structured as a non-linear relationship (that is, a power curve) between inventory levels and throughput. This technique supports the basic inventory principle that a distribution network with more inventory buffers — or nodes where you position inventory in the network to protect against fluctuating demand — will require more inventory than one with fewer buffers.

To complete this analysis, you typically rely on historical data, key performance indicators (KPIs) or general rules of thumb. However, this traditional approach has some limitations and can lead to problems when operating your new network down the road.

The more changes you’d like to make to your network, the less you can depend on a conventional approach

Historical data and KPIs may be adequate data inputs when the network changes you’re considering aren’t too different from your current network. More specifically, if you keep the same number of buffer echelons and have comparable lead-times from the sources of supply, then you’re probably ok using this approach.

If, however, you’re considering network changes that'd result in changes to lead time, lead time variability or variability in demand, then this conventional approach will fall short. Similarly, if your current network has only one distribution center (or none), or you want to see what effect ramping up inventory fill rates will have on required inventory levels and storage space, then this approach also will be misleading.

Use an optimization model to evaluate extensive network changes

A single optimization model is best if you’re evaluating substantial changes to your supply chain network while determining the right network configuration and inventory levels. The model must include supply chain constraints, capacities and costs, as well as acknowledge the impact of changes in lead-time, lead-time variability and demand variability. Designing your model in a way that considers all these components will help you avoid the “chicken and egg” sequential decision paradox.

It's critical to get inventory projections just right when designing your supply chain network. Inventory is one of the most effective, yet costly components in the entire supply chain. It can make your network more responsive to demand variability and less prone to shortages. But it also ties up significant working capital.

— By Jeff Zoroya