Most supply chain visibility tools show you lists. Lists of suppliers, lists of orders, lists of shipments. What they do not show you is the structure<\/em> \u2014 the web of dependencies that determines how a disruption at one node cascades through the entire system.<\/p>\n Network analysis, rooted in graph theory, gives you that structural view. It answers questions that no ERP report can:<\/p>\n R’s We model a supply chain with five tiers: raw material suppliers, component suppliers, manufacturers, distribution centers, and end markets. The relationships between them are directional \u2014 material flows downstream from raw suppliers to markets.<\/p>\n The result is a directed graph with 26 nodes and 41 edges spanning five tiers. Each edge represents a supply relationship, and the weight represents relative volume or importance.<\/p>\n The first step is to see the network as a whole. We color nodes by tier and size edges by volume:<\/p>\n The Sugiyama layout arranges nodes left-to-right by tier, so the material flow direction is immediately visible. Several structural patterns stand out:<\/p>\n Not all nodes in a supply chain are equally important. Betweenness centrality<\/strong> measures how many shortest paths between all pairs of nodes pass through a given node. A node with high betweenness is a chokepoint \u2014 if it fails, many supply paths are disrupted.<\/p>\n Mfg-3 dominates the betweenness ranking. It sits at the intersection of the most supply paths \u2014 five inbound connections from component suppliers and three outbound connections to distribution centers. This is your most critical node.<\/p>\n We can visualize this by scaling node size to betweenness:<\/p>\n The manufacturers visually dominate the map because they are the chokepoints. Distribution centers also show significant betweenness \u2014 they control access to end markets. Raw material suppliers and end markets have zero betweenness because they sit at the network edges.<\/p>\n A bar chart makes the ranking explicit:<\/p>\n For a procurement manager, this chart is a prioritized risk list. The nodes at the top are where you need backup suppliers, safety stock, or dual-sourcing strategies.<\/p>\n Degree centrality splits into two components in a directed network:<\/p>\n The manufacturers stand out with high values in both directions \u2014 they receive from many and supply to many. This dual dependency makes them critical from both the supply and demand side.<\/p>\n A node with high in-degree and low out-degree (like Market-2 or Market-3) is a demand concentrator \u2014 losing it means losing a significant share of revenue. A node with high out-degree and zero in-degree (the raw material suppliers) is a supply source \u2014 losing it affects everything downstream.<\/p>\n The real power of network analysis is simulation. We can remove the most critical node (Mfg-3) from the network and measure the impact:<\/p>\n The visual impact is immediate. Five component suppliers (Comp-2, Comp-3, Comp-5, Comp-6, Comp-7) lose one of their downstream paths. Some markets now have fewer supply routes reaching them.<\/p>\n The numbers quantify the damage:<\/p>\n Removing a single manufacturer eliminates 21% of all possible raw-to-market supply paths. This is the kind of insight that justifies dual-sourcing investments, strategic inventory buffers, or qualifying backup manufacturers.<\/p>\n In a real supply chain, you would run this simulation for every node with betweenness above a threshold. The result is a vulnerability matrix that tells you exactly which disruptions would hurt the most and where to invest in resilience.<\/p>\n Community detection algorithms identify groups of nodes that are more densely connected to each other than to the rest of the network. In a supply chain context, these clusters often represent natural supply ecosystems \u2014 groups of suppliers, manufacturers, and distributors that primarily interact with each other.<\/p>\n For procurement strategy, these clusters reveal:<\/p>\n This is directly actionable for supplier portfolio management. Clusters with single-source dependencies need diversification. Clusters that span all tiers need monitoring as integrated supply ecosystems.<\/p>\n Network analysis turns your supply chain from a list of suppliers into a strategic map. Here is what to do with these results:<\/p>\n 1. Protect the chokepoints.<\/strong> Nodes with high betweenness centrality (Mfg-3, Mfg-1, DC-1) are your single points of failure. Dual-source, buffer inventory, or qualify backup capacity at these locations.<\/p>\n 2. Run disruption scenarios quarterly.<\/strong> Remove each critical node from the graph and measure path loss. Compare the results against your current risk mitigation investments. If your highest-betweenness node has no backup plan, that is your top priority.<\/p>\n 3. Diversify within clusters.<\/strong> Community detection reveals where your supply base is concentrated. If an entire cluster depends on one raw material supplier, a single disruption cascades through the whole group.<\/p>\n 4. Monitor degree changes over time.<\/strong> If a supplier’s in-degree drops (they are losing their own suppliers), that is an early warning signal. If a manufacturer’s out-degree increases (they are taking on more customers), their capacity may become a constraint for you.<\/p>\n 5. Map your real network.<\/strong> The synthetic example here has 26 nodes. Your actual supply chain may have hundreds or thousands. The same igraph methods scale \u2014 betweenness centrality, community detection, and disruption simulation work on networks of any size. Start with your Tier 1 suppliers and expand upstream as data becomes available.<\/p>\n You need one CSV file: a list of supply relationships with The network is already there. You just have not drawn it yet.<\/p>\n\n
igraph<\/code> package makes this analysis accessible. In this post, we build a multi-tier supply chain network from scratch and apply the methods that reveal hidden vulnerabilities.<\/p>\nBuilding a Multi-Tier Supply Chain Network<\/h2>\n
library(igraph)\n\nset.seed(42)\n\n# Define the five tiers\nraw_suppliers <- paste0("Raw-", 1:6)\ncomp_suppliers <- paste0("Comp-", 1:8)\nmanufacturers <- paste0("Mfg-", 1:3)\ndistributors <- paste0("DC-", 1:4)\ncustomers <- paste0("Market-", 1:5)\n\n# Define supply relationships (edges)\nedges <- data.frame(\n from = c(\n "Raw-1","Raw-1","Raw-2","Raw-2","Raw-3","Raw-3",\n "Raw-4","Raw-4","Raw-5","Raw-5","Raw-6","Raw-6",\n "Comp-1","Comp-1","Comp-2","Comp-2","Comp-3","Comp-3",\n "Comp-4","Comp-5","Comp-5","Comp-6","Comp-7","Comp-8",\n "Comp-4","Comp-7",\n "Mfg-1","Mfg-1","Mfg-2","Mfg-2","Mfg-3","Mfg-3","Mfg-3",\n "DC-1","DC-1","DC-2","DC-2","DC-3","DC-3","DC-4","DC-4"\n ),\n to = c(\n "Comp-1","Comp-2","Comp-2","Comp-3","Comp-4","Comp-5",\n "Comp-5","Comp-6","Comp-7","Comp-8","Comp-8","Comp-1",\n "Mfg-1","Mfg-2","Mfg-1","Mfg-3","Mfg-2","Mfg-3",\n "Mfg-1","Mfg-2","Mfg-3","Mfg-3","Mfg-1","Mfg-2",\n "Mfg-2","Mfg-3",\n "DC-1","DC-2","DC-2","DC-3","DC-3","DC-4","DC-1",\n "Market-1","Market-2","Market-2","Market-3",\n "Market-3","Market-4","Market-4","Market-5"\n )\n)\n\nedges$weight <- sample(10:100, nrow(edges), replace = TRUE)\n\nall_nodes <- c(raw_suppliers, comp_suppliers, manufacturers,\n distributors, customers)\n\ng <- graph_from_data_frame(edges, directed = TRUE,\n vertices = data.frame(name = all_nodes))\n<\/code><\/pre>\nVisualizing the Network Structure<\/h2>\n
V(g)$tier <- ifelse(grepl("Raw", V(g)$name), "Raw Material",\n ifelse(grepl("Comp", V(g)$name), "Component",\n ifelse(grepl("Mfg", V(g)$name), "Manufacturer",\n ifelse(grepl("DC", V(g)$name), "Distribution",\n "Market"))))\n\ntier_colors <- c("Raw Material" = "#2196F3", "Component" = "#4CAF50",\n "Manufacturer" = "#FF9800", "Distribution" = "#9C27B0",\n "Market" = "#F44336")\n\nV(g)$color <- tier_colors[V(g)$tier]\n\nplot(g,\n layout = layout_with_sugiyama(g, layers = tier_order)$layout,\n vertex.size = 18,\n vertex.color = V(g)$color,\n edge.arrow.size = 0.4,\n edge.width = E(g)$weight \/ 30,\n main = "Multi-Tier Supply Chain Network")\n<\/code><\/pre>\n![\"Multi-Tier](\"https:\/\/inphronesys.com\/wp-content\/uploads\/2026\/02\/network_full_scm.png\")
<\/p>\n\n
Finding Critical Nodes with Centrality Analysis<\/h2>\n
bet <- betweenness(g, directed = TRUE)\ndeg <- degree(g, mode = "all")\n\ncentrality_df <- data.frame(\n node = V(g)$name,\n tier = V(g)$tier,\n degree = deg,\n betweenness = round(bet, 1)\n)\n\ncentrality_df[order(-centrality_df$betweenness), ]\n<\/code><\/pre>\n node tier degree betweenness\n Mfg-3 Manufacturer 8 58.0\n Mfg-1 Manufacturer 6 33.0\n DC-1 Distribution 4 29.0\n Mfg-2 Manufacturer 7 24.0\n DC-3 Distribution 4 20.0\n Comp-1 Component 4 18.0\n DC-2 Distribution 4 16.0\n Comp-7 Component 3 11.0\n DC-4 Distribution 3 11.0\n Comp-2 Component 4 9.0\n<\/code><\/pre>\nnode_size <- 10 + (bet \/ max(bet)) * 25\n\nplot(g, layout = layout,\n vertex.size = node_size,\n vertex.color = V(g)$color,\n main = "Betweenness Centrality \u2014 Node Size = Criticality")\n<\/code><\/pre>\n![\"Betweenness](\"https:\/\/inphronesys.com\/wp-content\/uploads\/2026\/02\/network_centrality.png\")
<\/p>\nlibrary(ggplot2)\n\ntop_10 <- centrality_df[order(-centrality_df$betweenness), ][1:10, ]\ntop_10$node <- factor(top_10$node, levels = rev(top_10$node))\n\nggplot(top_10, aes(x = node, y = betweenness, fill = tier)) +\n geom_col(width = 0.7) +\n coord_flip() +\n scale_fill_manual(values = tier_colors) +\n labs(title = "Top 10 Nodes by Betweenness Centrality",\n subtitle = "Higher betweenness = more supply paths depend on this node",\n x = NULL, y = "Betweenness Centrality")\n<\/code><\/pre>\n![\"Top](\"https:\/\/inphronesys.com\/wp-content\/uploads\/2026\/02\/network_betweenness_bar.png\")
<\/p>\nIn-Degree and Out-Degree: Who Depends on Whom?<\/h2>\n
\n
deg_df <- data.frame(\n node = V(g)$name,\n tier = V(g)$tier,\n in_degree = degree(g, mode = "in"),\n out_degree = degree(g, mode = "out")\n)\n\nggplot(deg_df, aes(x = reorder(node, in_degree + out_degree), fill = tier)) +\n geom_col(aes(y = out_degree), width = 0.7, alpha = 0.9) +\n geom_col(aes(y = -in_degree), width = 0.7, alpha = 0.6) +\n coord_flip() +\n scale_fill_manual(values = tier_colors) +\n labs(title = "In-Degree and Out-Degree by Node",\n subtitle = "Right = supplies to | Left = receives from",\n x = NULL, y = "\\u2190 In-Degree | Out-Degree \\u2192")\n<\/code><\/pre>\n![\"In-Degree](\"https:\/\/inphronesys.com\/wp-content\/uploads\/2026\/02\/network_degree_dist.png\")
<\/p>\nDisruption Simulation: What Happens When a Critical Node Fails?<\/h2>\n
critical_node <- "Mfg-3" # Highest betweenness\n\ng_disrupted <- delete_vertices(g, critical_node)\n\nplot(g_disrupted,\n layout = layout_with_sugiyama(g_disrupted, layers = layer)$layout,\n vertex.size = 18,\n vertex.color = V(g_disrupted)$color,\n main = paste0("Disruption Scenario: ", critical_node, " Removed"))\n<\/code><\/pre>\n![\"Disruption](\"https:\/\/inphronesys.com\/wp-content\/uploads\/2026\/02\/network_disrupted.png\")
<\/p>\nRaw-to-Market paths before disruption: 29\nRaw-to-Market paths after disruption: 23\nPaths lost: 6\n<\/code><\/pre>\nCommunity Detection: Finding Natural Clusters<\/h2>\n
g_undirected <- as_undirected(g, mode = "collapse")\ncommunities <- cluster_louvain(g_undirected)\n\nplot(g,\n layout = layout,\n vertex.size = 18,\n vertex.color = community_colors[membership(communities)],\n main = paste0("Community Detection (Louvain) \u2014 ",\n max(membership(communities)), " Clusters"))\n<\/code><\/pre>\n![\"Community](\"https:\/\/inphronesys.com\/wp-content\/uploads\/2026\/02\/network_communities.png\")
<\/p>\nCluster 1: Raw-1, Comp-1, Comp-2, Mfg-1, Mfg-3, DC-1, Market-1\nCluster 2: Raw-2, Comp-3\nCluster 3: Raw-3, Raw-4, Comp-4, Comp-5, Comp-6, Mfg-2\nCluster 4: Raw-5, Raw-6, Comp-7, Comp-8\nCluster 5: DC-2, Market-2\nCluster 6: DC-3, DC-4, Market-3, Market-4, Market-5\n<\/code><\/pre>\n\n
From Analysis to Action<\/h2>\n
Getting Started<\/h2>\n
# Install igraph\ninstall.packages("igraph")\nlibrary(igraph)\n\n# Load your own supplier relationship data\nedges <- read.csv("supplier_relationships.csv") # columns: from, to, volume\ng <- graph_from_data_frame(edges, directed = TRUE)\n\n# Immediate analysis\nbetweenness(g, directed = TRUE) # Find chokepoints\ndegree(g, mode = "all") # Find most connected nodes\ncluster_louvain(as_undirected(g)) # Find natural clusters\n\n# Visualize\nplot(g, vertex.size = 10 + betweenness(g) \/ max(betweenness(g)) * 20)\n<\/code><\/pre>\nfrom<\/code> and to<\/code> columns. Every ERP system can export this. Your purchasing records already contain the data \u2014 each purchase order connects a supplier to your company, and your BOMs connect components to finished goods.<\/p>\nInteractive Dashboard<\/h2>\n