The Sterling Simulation consulting company was hired by a pharmaceutical giant to create a supply chain model. The client was introducing one of their products to new markets and wanted to visualize the restructuring of the supply chain and how it would react to demand uncertainty. The company’s goal was to reduce lead time while maintaining fill rate and avoiding backorders or lost sales. The client wanted a flexible ‘what if’ tool for risk mitigation analysis.
The client had previously built a discrete-event supply chain model in AnyLogic. However, it lacked flexibility because it could not be easily adjusted to new conditions. Sterling Simulation engineers offered a solution: build a hybrid model, which would combine agent-based supply chain components with discrete-event processes. The new model was easier to configure and use, as it allowed for run-time construction of both the supply chain itself and the processes inside each component.
The model’s core included agents that represented:
- Production facilities containing one or more processes and shared resources.
- Production lines, which incorporated production and shipping logic.
- Process steps, representing single steps in production process.
- Types of batches with preassigned sizes and required raw materials.
- Shipment requests for product batches.
Production and shipment processes in the new model were initially based on current MRP scheduling. After the initial processing, the model used MRP scheduling to determine production needs. The model was impacted by demand variability as it created surpluses or shortfalls in safety stocks. This allowed the client’s prime goals of satisfying demand and reducing financial and operational risks to be met organically by the model.
To determine whether to produce, the model calculated the sum of current demand and desired safety stock. If it was less than current inventory, they did not produce, otherwise they charged inventories and started production.
To validate the model, the client could not use the existing supply chain because it did not include the new markets. By leveraging the model’s configurability, they applied it to a different supply chain which had available data, and showed that the model's logic worked properly.
The model’s hybrid nature allowed analysts to combine supply chain components, including production facilities and lines, with features of production and shipment flows, making the model highly flexible. It was largely possible due to AnyLogic’s hybrid modeling abilities, which combined agent based and discrete-event approaches. Thanks to AnyLogic, model building time was also drastically reduced. It also allowed for natural minimalist design, easier testing, and verification.
The model showcased how lead time reduction techniques might be implemented in a pharmaceutical supply chain, which could reduce inventory.
To include more dynamics into the supply chain model, and conduct in-depth experiments with it, the company and the contractor want to shift the AnyLogic model to anyLogistix, a specialized software for supply chain design and analytics. This approach could bring more insights to the company’s executives, including the case of supply chain expansion.