Chart Pattern · Reference

Network Graph

Nodes + edges · entity relationships

Customer–supplier–partner network

Each node = entity (size = revenue) · each edge = a transactional relationship · clusters reveal sub-networks

✓ When to use

  • Showing relationships between entities (orgs, people, fraud rings)
  • Identifying hubs and peripheral nodes
  • Detecting clusters or communities in the data

✗ Avoid when

  • You really want a flow (volume between stages) — Sankey is better
  • You have >100 nodes — graph becomes a hairball; use a matrix
  • Hierarchical relationships only — use a tree diagram

★ Best practice

  • Size by importance (revenue, degree centrality)
  • Color by cluster or category
  • Lay out with a force-directed algorithm — never random
  • Label only the most important nodes; let the eye discover the rest