Jinwoo Je

Introduction

Maintaining optimal inventory levels across a complex global supply chain network presents a significant challenge, particularly in industries where product availability is critical. Disruptions in supply and demand can quickly lead to costly imbalances, resulting in excess stock prone to obsolescence and deficits leading to backorders and lost sales. Boston Scientific (BSC), a global leader in the medical device industry operating in over 120 countries, faced precisely this challenge, experiencing inventory imbalances across its distribution network, exacerbated by the Covid-19 pandemic. This project explored a powerful solution: inventory rebalancing through lateral transshipment, the strategic repositioning of stock between distribution centers (DCs) at the same echelon. The objective was to develop and validate a methodology that could optimize inventory positions, reduce costs, and enhance service levels across BSC's extensive network.

The Challenge: Inventory Imbalance at BSC

BSC manages a vast supply chain, encompassing manufacturing sites, sterilization facilities, and a multi-tiered distribution network designed to serve thousands of hospitals worldwide. Despite sophisticated planning systems like Multi-Echelon Inventory Optimization (MEIO) and a "Healthy Stock Model" to define acceptable inventory ranges, significant imbalances persisted. Analysis revealed that while 36% of inventory was in excess, 9% was in deficit across different nodes. BSC's existing network structure allowed only unidirectional product flow (manufacturing to Tier 1 DC to Tier 2 DC), preventing the repositioning of excess stock to cover deficits in other locations. This rigidity meant excess inventory often resulted in scrap costs upon expiration, while deficits led to backorders and potential lost sales. The fluctuating demand patterns during and after the Covid-19 pandemic further intensified these imbalances, creating a clear need for a more flexible inventory management strategy.

Methodology: Modeling and Optimization

This study employed a rigorous, data-driven approach centered on Mixed Integer Linear Programming (MILP) to determine optimal transshipment solutions. Data encompassing inventory levels, forecasts, safety stock parameters, lead times, and costs were extracted primarily from BSC's ERP systems (SAP and Rapid Response). Transportation costs for potential new transshipment lanes were estimated using external tools like the UPS CTC calculator, acknowledging the lack of pre-existing routes for lateral movements.

The core of the methodology involved developing and scaling MILP optimization models. Initial models focused on minimizing transportation costs for single or small batches of SKUs. More complex models were then developed, incorporating not only transportation costs but also the costs associated with holding excess inventory (potential scrap) and experiencing deficits (backorders/lost sales). These models aimed to find the optimal balance between the cost of transshipping goods and the cost savings from reducing excess and deficit inventory levels.

Recognizing the computational complexity of analyzing thousands of SKUs (BSC manages over 16,000 finished goods SKUs), two primary strategies were used for selecting high-priority SKUs for the full-scale optimization:

ABC Classification: Leveraging BSC's existing revenue-based SKU categorization (A-Highest, B-Medium, C-Lowest revenue) to select the top 1,000 SKUs.
Potential Balance Opportunity (PBO) Heuristic: A novel heuristic developed for this project, calculating the potential savings for each SKU by identifying the minimum value between its total excess and total deficit across all DCs. The top 1,000 SKUs based on PBO ranking were selected.

Finally, Monte Carlo simulations were conducted to evaluate the performance of the rebalanced inventory model compared to the baseline (no transshipment) under conditions of stochastic demand, using moderate and high demand variability scenarios based on historical data.

Results: Significant Cost Savings and Improved Inventory Health

The optimization results demonstrated the substantial benefits of implementing lateral transshipment. The MILP models consistently identified transshipment strategies that reduced overall inventory costs.

For the model using ABC-selected SKUs (Model 4a), optimization resulted in a 10.2% reduction in total inventory costs, decreasing from $34.6M to $31.1M. This involved executing 1,666 transshipments at a cost of $2.6M, which successfully reduced excess inventory costs by $3.3M and deficit costs by $2.8M. Simulation studies further showed that under both moderate and high demand variability, the rebalanced state maintained a higher percentage of SKUs within the desired "Healthy Stock Zone" compared to the base case (e.g., 43.2% vs 37.1% in moderate CV) and significantly reduced the amount of excess inventory.

The model utilizing the PBO heuristic for SKU selection (Model 4b) yielded even more impressive results. This approach led to a 25% reduction in total costs, from $87M to $65M. While incurring higher transshipment costs ($7.3M for 3,681 transshipments), the savings generated from reducing excess ($15.2M reduction) and deficit ($13.9M reduction) inventory were substantially greater. Simulation results again confirmed superior performance, with the rebalanced model achieving a significantly higher percentage of SKUs in the healthy zone (e.g., 47.7% vs 34.8% in high CV) and drastically reducing excess inventory levels.

Across both models, the primary transshipment flows occurred logically between major regional hubs, notably between Europe and Asia, and between North America and Latin America. These results strongly indicate that lateral transshipment offers a viable and highly effective strategy for mitigating inventory imbalances and associated costs.

Discussion and Implementation Roadmap

The study confirms that lateral transshipment can unlock significant savings, potentially between 10% and 25% of total inventory costs for the analyzed SKUs. To translate these findings into practice, a phased implementation approach is recommended. BSC could initiate a pilot project focusing on SKUs selected via the existing ABC classification, leveraging familiar internal metrics. Success in the pilot phase could justify scaling the program and adopting more sophisticated selection methods like the PBO heuristic, which demonstrated higher savings potential.

Integrating the rebalancing analysis into BSC's existing periodic inventory review process would streamline implementation, minimizing the need for additional dedicated resources. Furthermore, embedding the monitoring of imbalance levels and the execution of transshipments within BSC's existing ERP systems would be crucial for user adoption and operational efficiency. Collaborating with logistics partners to optimize transportation routes and consolidate volumes could further enhance cost-effectiveness beyond the direct point-to-point costs modeled in this study.

It is important to acknowledge the study's limitations. The current models did not include intra-DC handling costs (picking, packing, receiving) associated with transshipment or potential fixed costs for setting up new shipping lanes. Additionally, complex global regulations governing medical device movement, which could restrict certain transshipments, were beyond the project scope and require careful consideration during implementation.

Conclusion

This project successfully demonstrated that inventory rebalancing through lateral transshipment is a powerful strategy for addressing inventory imbalances within BSC's complex distribution network. By strategically repositioning stock between DCs, BSC can significantly reduce costs associated with excess inventory and stockouts, achieving savings of up to 25%. The optimized inventory model proved robust, maintaining superior inventory health even under stochastic demand conditions. This study provides BSC with a clear methodology and compelling evidence to pursue lateral transshipment, contributing valuable insights applicable to other companies facing similar supply chain challenges. Further research could explore the dynamic interaction between replenishment cycles and transshipment activities and the applicability of these strategies in industries with different margin structures and demand profiles.