Christian D. Ruf shares a detailed article on using prescriptive analytics to design a supply chain network.
One of the duties I frequently performed as an operations research analyst in consulting projects was optimizing companies’ supply chain network designs. A supply chain is a network that connects suppliers with customers to procure materials, transform them into final products, and deliver these products to customers. Supply chain management is a key function of most companies and one of the most exciting management areas. Its goal is to efficiently match supply and demand for final products or services by designing and operating the supply chain. At the strategic level, supply chain design structures the supply chain to strike the right balance between procurement, inventory, transportation, and manufacturing costs as well as the strategic fit. The first step in this process is to determine the supply chain network design, which involves two major decisions:
Location of sites including production facilities, distribution centers, fulfillment centers, and stores
Should we add sites to the network?
Should we shut down existing sites?
Flow of products
From where should we source products, and how much should we produce and procure?
How should we move our products through the network, i.e. which routes and which modes of transport should we use?
These are strategic decisions with significant impact on the company’s cost structure, the resilience as well as the flexibility of the supply chain, and the speed at which products can be shipped to customers. While there is still an inclination to go with the gut when looking at an array of possible supply chain designs, advanced companies understand the value of optimization models to foster a rational decision-making process (see, e.g. Kuttappa, 2020; Gartner, 2019; Burtch Works, 2019; MarketWatch, 2021; Marker, 2017).
There exist an array of technologies for such prescriptive models: open source technologies such as PuLP for Python, SolverStudio for Excel, lpSolve for R, and google’s OR-Tools, as well as powerful commercial technologies such as Gurobi and CPLEX. Additionally, there are vendors offering solutions for common supply chain models, which can potentially increase the productivity to some extent. However, they are less flexible than custom built models and can hence require cumbersome workarounds when there are requirements that are not captured in the standard models.
All network design optimization models are extensions of the well-known facility location problem (or warehouse location problem). Therefore, this post outlines the facility location problem and some scenarios that are typically analyzed during supply chain network design projects. It also elaborates typical ways in which this model is extended to capture requirements of real network design projects.
Key points include:
- The facility location problem
- The baseline scenario
- The optimized baseline scenario
Read the full article, Prescriptive analytics for supply chain network design, on CDRuff.com.