The problem with multi-period network design for seasonal products — and how to fix it

  • May 19, 2022
Watermelon slice popsicles on a blue rustic wood background

If your company manufactures seasonal items, then you know that these products often add complexity to distribution network planning. The answer for many companies is to use multi-period network design models to understand their seasonal storage requirements.

In theory, by representing the time-phased imbalance between supply and demand, a network design model will calculate the expected storage requirements. Presumably by building inventory in advance of the imbalance. However, the theoretical and the realistic operate on different planes.

Modeling seasonal storage requirements requires a high level of detail. You must be confident in your build-ahead inventory profile, from both a time and location perspective. But that profile is often at a level too granular to be consistent with the scope of an overall network design project. In the real world, it takes multiple factors to establish the appropriate level of build-ahead inventory. For example, complex and overlapping production capacity constraints, batch size requirements and planning heuristics. Add in the manufacturing and operational realities of safety stock and target inventory requirements, transportation capacity and transportation load building requirements, and the model becomes incredibly complex.

In truth, reflecting all the constraints necessary to produce an intuitive build-ahead inventory profile often requires more effort than is realistically possible in a network design model. Mixed-integer optimization applications view the world in simpler terms than production managers or production and scheduling systems, despite advancements in enabling technology.

For example, historical time-phased inventory behavior would be challenging to replicate based solely on the imbalance between supply (that is, production) and demand.

So, even when you build a network design model that contains most of the real-world supply constraints, the time-phased answers can still lack credibility — especially at a location level. Why? Because the math the application uses to solve the problem is vastly different than the rules used to create operational plans.

However, when business conditions warrant, an alternative approach is available. In these situations, production planning provides a location level, time-phased inventory plan. The plan may already exist, either in support of sales and operations planning or as rules of thumb for distribution planning. Either way, you can incorporate these plans or rules into a network model as time-phased constraints. Doing so in conjunction with a handful of other appropriate constraints increases confidence among operations teams. And that increases the likelihood of implementing the results, which accelerates the time to actual value.

By Jeff Zoroya

Subscribe to our blog


Related Blog Posts