SPORD: A Simulation-Propose-then-OR-Dispose Approach for Supply Chain Planning introduces a new methodology to solve the complex, fragmented, and often slow planning processes currently used in large-scale e-commerce supply chains. By decoupling the evaluation of business rules from the final optimization, the authors provide a unified framework that allows companies to move from static, manual planning to a dynamic, automated system that is both computationally efficient and transparent to stakeholders.
The Core Problem: Why Planning Fails
Supply chain planning at large firms has historically been plagued by three major barriers. First, "operational fragmentation" means that because every business unit has unique rules, analysts must build bespoke models from scratch for every task, which is time-consuming and difficult to maintain. Second, "computational intractability" arises because the sheer scale of modern supply chains—involving millions of products and complex routing—makes it impossible for traditional solvers to find an optimal solution quickly. Finally, there is an "implementation hurdle": executives often distrust "black-box" mathematical models, making it difficult to adopt the solutions even when they are technically optimal.
How SPORD Works
The SPORD method addresses these issues through a "Simulation-Propose-then-OR-Dispose" architecture. Instead of trying to force complex, non-linear business rules into a rigid mathematical formula, the system uses a two-step process: 1. Simulation-Propose: The system uses a massively parallel simulation engine to generate and evaluate all "operationally valid" candidate paths. This step absorbs all the idiosyncratic business logic—such as storage rules or routing constraints—that would normally break a standard optimization model. 2. OR-Dispose: Once the simulation has evaluated the costs and service impacts of these valid paths, an integer programming model (the "OR" part) selects the globally optimal subset.
By separating the "what is possible" (simulation) from the "what is best" (optimization), the system becomes modular and easier to configure for different business needs without requiring a total reconstruction of the model.
Computational and Operational Impact
To handle the massive scale of data, the authors implemented a matrix-vectorized simulation engine that runs on CPUs and GPUs, achieving speeds 10 to 100 times faster than traditional serial methods. They also utilized a list-scheduling algorithm to manage complex order dependencies in minutes rather than hours.
The method has been deployed as the "NetSim" platform at JD.com, where it has been used to optimize services for over 20,000 suppliers. Since 2025, the platform has demonstrated significant real-world results: the cross-regional fulfillment rate improved from 6.1% to 4.9%, and the system has achieved an average monthly carbon reduction of approximately 5,745 tCO2e.
Building Trust Through Transparency
A key advantage of the SPORD approach is its ability to turn skeptical executives into collaborators. Because the simulation engine provides transparent, verifiable outputs based on actual supply chain "physics" and business rules, stakeholders can see exactly why a specific path was chosen. This allows for "what-if" analysis, where executives can compare the proposed solution against alternatives, fostering the trust necessary to move from theoretical planning to active, real-world implementation.
Comments (0)
to join the discussion
No comments yet
Be the first to share your thoughts!