Status
Usable for directional analysis, not positioned as a finished optimization engine.
Case study
Building a Graph-Based Warehouse Optimization and Simulation Workflow
A warehouse graph model and simplified simulation made warehouse optimization ideas testable before building a full optimization engine.
Published Mar 23, 2026 Updated Mar 24, 2026
Overview
Warehouse optimization started with building a graph the team could actually analyze.
Starting point
We had storage node information such as coordinates, but no practical graph that represented how workers could actually move through the warehouse. That made it difficult to reason about layout, access, inventory placement, or picker travel distance in a structured way.
The immediate problem was not optimization yet. It was building the path-aware model needed before any meaningful simulation or routing analysis could happen.
Current scope
I am not presenting this as a finished warehouse optimization engine. The current system has two parts: a tool for constructing the warehouse graph, and a simplified simulation model for estimating how inventory reallocation changes picker travel distance.
The goal is directional decision support, not a perfect digital twin.
Graph construction
Turn storage nodes and walkable paths into an analyzable warehouse graph.
The graph comes first. Once movement is represented cleanly, routing and simulation become possible on top of it.
Node grouping
Storage nodes are grouped first. The current approach uses Manhattan distance to form practical groups before routing is computed at a larger scale.
Access points
Each group is assigned one or more access points so the graph can represent how workers enter and interact with that part of the warehouse.
Path-based connections
Users draw walkable paths, and the tool automatically connects nearby access points to those paths to generate a usable graph structure.
Why grouping matters
Running shortest-path calculations directly across every storage node would be expensive. Grouping reduces graph complexity so later routing and simulation work remains tractable.
Flow
How the warehouse model is constructed
Simulation and routing only become useful after the warehouse can be represented as a usable movement model.
Simulation
A simplified model for testing inventory reallocation ideas.
Rule-based ranking
Products and storage nodes are ranked using a simplified grading logic instead of a full operational model.
Reallocation test
Inventory is reassigned according to those grades, and the model compares estimated picker travel distance before and after reallocation.
Directional output
The result is meant to explain whether better placement should materially reduce movement, not to claim an exact final operating result.
Results
The number is useful as a signal, not as a final claim.
Estimated
~50%Travel distance reduction in the simplified simulation.
Interpretation
Not finalThe result is directional and does not fully reflect real warehouse constraints.
Key risk
BottlenecksLocal congestion may appear if picking activity concentrates around specific locations.
Takeaways
The important output is structured explanation, not just a number.
Signal
Even a simplified model can show that inventory placement changes may have a meaningful effect on travel distance.
Limit
The current simulation does not fully model congestion or operational constraints, so the 50 percent figure should not be treated as a final operating result.
Related
Continue with connected pages
Case studies and methods that connect to the same operational questions.
Case study
AI-Powered Text-to-SQL System
Built an AI-powered Text-to-SQL system that used structured metadata, Athena execution, and validation loops to speed up ad-hoc analytics.
Method
Seasonality Index for Inventory Planning
An inventory planning method that estimates SKU-level seasonality, groups products into seasonal buckets, and connects those signals back to category planning.