Optimizing supply chain resilience and decrease operational costs with data-driven insights
Challenge
Achieving efficiency in safety stock and order sizing
Managing an efficient supply chain is a delicate balance between ensuring product availability and minimizing excess inventory. Our client, a major player in the pharmaceutical industry, faced a common but complex challenge: their safety stock levels were too high, leading to increased costs and inefficiencies. At the same time, strict regulatory requirements made stockouts non-negotiable. They sought a data-driven approach to determine the optimal safety stock and lot size while maintaining the required service levels.
Approach
Combining simulation and Machine Learning for optimized decision-making
To fully understand the existing supply chain dynamics, we began by interviewing key personnel involved in the process. This provided a clear overview of how materials flowed through the system, revealing inefficiencies and bottlenecks. We then moved on to data collection, gathering all available information on demand patterns, lead times, and stock levels. While sufficient data was available, data quality posed a challenge that had to be addressed in our modeling efforts.
We explored two methodologies to optimize safety stock:
- Monte Carlo Simulations – A probabilistic approach to simulate demand and supply variability, allowing us to determine the optimal safety stock levels.
- Machine Learning-Based Forecasting – Leveraging historical data to predict demand fluctuations and improve stock calculations.
While the ML approach showed promise, the available data was insufficient for building an accurate predictive model. Orders were placed on irregular timeframes (weekly, monthly, or even longer intervals), limiting the granularity required for effective forecasting. Given these constraints, we focused on the Monte Carlo simulation approach.
By modelling demand distributions and supply chain variability, we ran 100,000 simulations per material, identifying the necessary safety stock levels to meet the client's desired service levels. We then validated our recommendations against historical supply chain data to ensure no stockouts would occur under the revised safety stock levels.
Impact
Reducing safety stock by over 40% without sacrificing reliability
The results were significant. Our simulation-driven approach led to a 41% reduction in overall safety stock levels while maintaining the required service levels. This optimization not only reduced excess inventory but also improved supply chain efficiency and lowered holding costs.
Beyond immediate results, we provided recommendations on improving data collection processes to enhance future ML-driven supply chain optimizations. By refining data quality and increasing the granularity of demand tracking, our client could further leverage AI-driven forecasting in future iterations of their supply chain strategy.
This project demonstrated the power of advanced analytics in supply chain management. By combining deep domain knowledge with data-driven methodologies and a hint of Machine Learning, we successfully optimized stock levels while ensuring compliance with stringent regulatory requirements. The outcome? A leaner, more resilient supply chain ready to adapt to future challenges. Oh yes, and cost savings. That’s right.