network-design

What it does

The skill directs the agent to design or optimize a supply chain network by choosing facility locations, network echelons, and flow strategies, and to explore hub-and-spoke concepts and distribution strategies. It treats network design as balancing facility, transportation, inventory, and service costs. It explains types of facilities (manufacturing plants, DCs, cross-docks, forward stocking locations, 3PLs) and flow strategies (Make-to-stock, Make-to-order, Configure-to-order, Drop ship). It also covers high-level optimization problem types (Uncapacitated Facility Location, Capacitated Facility Location, P-Median, P-Center, Hub Location) and provides a Mixed-Integer Programming model with decision variables, objective, and constraints, plus a PuLP-based Python implementation skeleton for solving the network design problem. It includes sections on cost components (Fixed Facility Costs, Transportation Costs, Inventory Costs, Service Level Metrics) and a step-by-step network design process (Data Collection & Validation, Network Strategy Definition, Scenario Modeling, Optimization & Trade-off Analysis, Recommendation & Implementation Planning). It lists tools and techniques, including optimization software and Python network design code examples, and demonstrates a multi-scenario optimization workflow. It instructs how to prepare data (customers, facilities, demand, costs) and how to run optimizations to obtain opened facilities and flow allocations.

How it works

• Begin with an Initial Assessment gathering Business Context, Strategic Objectives, Data Availability, and Constraints.
• Use the Network Design Framework to decide Strategic Network Decisions: Number of Echelons, Facility Types, and Flow Strategies.
• Apply Network Optimization Models for Facility Location (UFL, CFL, P-Median, P-Center, Hub Location).
• Use the Mathematical Optimization Approach (Mixed-Integer Programming) with variables y_j (binary, facility opened) and x_ij (flow from facility to customer). Define Objective to minimize fixed facility costs, transportation costs, handling costs, and inventory costs. Enforce constraints: demand satisfaction, facility capacity, and that flows exist only from opened facilities.
• Provide a Python Implementation snippet using PuLP that sets up the problem, defines decision variables, builds the objective and constraints, solves, and extracts results like total_cost, open_facilities, and flows.
• Outline Cost Components (Fixed Facility Costs, Transportation Costs, Inventory Costs, Service Level Metrics) with typical ranges and simple formulas (e.g., Square Root Law for safety stock).
• Describe the Network Design Process in phases: Data Collection, Strategy Definition, Scenario Modeling, Optimization & Trade-off Analysis (including a multi-scenario loop and Pareto frontier), and finally Recommendation & Implementation Planning.
• List Tools & Techniques including commercial solvers (LLamasoft, Gurobi, CPLEX, etc.) and Open Source options (PuLP, Pyomo, OR-Tools, JuMP). It provides a Python network design example skeleton and distance calculations to support modeling.

When to use it

Use when the user wants to design or optimize a supply chain network, determine facility locations, or configure distribution strategies. It is also applicable when the user mentions network optimization, facility location, DC location, distribution network, greenfield analysis, brownfield optimization, or hub-and-spoke. For transportation routing within an existing network, refer to route-optimization. The skill supports exploring base, optimized current, greenfield, and growth scenarios with sensitivity analyses and multi-scenario optimization.

What it can touch

• Open facilities (binary decisions) and flows (continuous decisions) as part of the MIP model.
• It references data inputs like customer locations, demand, facility costs, capacities, and transportation/handling costs to drive the optimization.

Caveats

• The content is framed around a modeling approach and example code; real-world results depend on data accuracy and solver availability. It presents typical cost components, ranges, and modeling assumptions, but does not guarantee outcomes.