Context Graphs: Seeing Your Supply Chain as It Actually Is

Picture this: a fire breaks out at a semiconductor fab in Taiwan. Within hours, automotive production lines in Germany start idling. A week later, a dishwasher factory in Ohio can't fulfill orders. Two months later, your company misses quarterly targets because you couldn't ship products that needed a $3 chip.

This isn't hypothetical. It happened in 2021. And most companies didn't see it coming until it was too late.

The problem wasn't a lack of data. Companies had spreadsheets full of supplier information. They had ERP systems tracking orders and inventory. What they didn't have was a way to see how everything connected—and what would happen when one connection broke.

Your Supply Chain Is a Graph

Think about your supply chain for a moment. You have suppliers. Those suppliers have suppliers. Raw materials flow from mines to refineries to component manufacturers to assembly plants to distribution centers to stores. At every node, things transform: aluminum becomes casings, silicon becomes chips, chips become modules, modules become products.

This isn't a list. It's a graph.

A context graph captures these relationships explicitly. Node A supplies Node B. Node B transforms inputs into outputs. Node C depends on Node B's outputs. When something happens at Node A, we can trace exactly who gets affected downstream.

Toyota figured this out decades ago. After the 2011 earthquake and tsunami in Japan, they spent two years mapping their supply chain down to the raw material level. They discovered they had over 400,000 suppliers across multiple tiers. More importantly, they built a system to track how disruptions would propagate through that network.

When the 2016 Kumamoto earthquake hit, Toyota recovered production in two weeks instead of months. They knew exactly which suppliers were affected and which backup paths existed.

Events and Entities

A context graph has two core concepts: entities and events.

Entities are the things in your supply chain. Suppliers, plants, warehouses, carriers, ports, customers. Each entity has attributes—location, capacity, lead times, the products it handles.

Events are what happens to those entities. An order placed. A shipment dispatched. A customs hold. A quality inspection failed. A supplier declares force majeure.

Here's where it gets interesting: events flow through the graph.

When a container ship runs aground in the Suez Canal (March 2021, the Ever Given), that's an event at one node. But that event creates cascading events downstream. Delayed arrivals at European ports. Missed manufacturing windows. Backlogged distribution centers. Each delay creates new delays.

A context graph lets you trace these cascades before they happen. You can ask: if this ship is delayed by 6 days, which orders miss their delivery windows? Which production runs get pushed? Which customers need to be notified?

Cause and Effect: The Missing Link

Here's what most supply chain systems miss: they track events, but they don't understand why events happen.

Your ERP tells you an order is late. But why is it late? Was the shipment delayed? Did the supplier miss their production window? Did a quality check fail upstream? Did a raw material shipment get held at customs three weeks ago? Without causal relationships encoded in your graph, you're stuck playing detective every time something goes wrong—exactly the problem automated root cause analysis is designed to solve.

A context graph captures not just what happened, but what caused it.

Consider a real scenario: a major retailer notices their inventory of a popular toy is running low in October, right before the holiday season. Their system shows the supplier shipped on time. The distributor delivered on time. So where's the problem?

It turns out a resin shortage in South Korea three months earlier forced a plastics manufacturer in Vietnam to reduce output. That manufacturer supplies a component maker in China, who supplies the toy factory in Mexico. By the time the shortage hit the toy factory, nobody could trace it back to the original cause. They were all just reacting to symptoms.

With causal relationships modeled, you can trace backward from effect to cause. Late delivery → missed production slot → component shortage → supplier delay → raw material disruption. Each link in the chain is explicit. When you see the resin shortage happen, you can predict the toy shortage months before it hits.

This works in both directions. Trace forward: what will this event cause? Trace backward: what caused this event? Both questions need answers if you want to respond intelligently.

The 2020 toilet paper shortage is a famous example. Demand spiked 700% in a week. But the supply chain didn't fail because of production capacity. It failed because distribution networks were optimized for steady commercial and retail channels. When consumers suddenly bought ten times their normal amount at retail, the system couldn't reroute commercial inventory fast enough. The cause wasn't shortage. It was a demand pattern shift that the distribution graph couldn't handle. Companies that understood this causal structure were able to adapt their routing. Those that didn't just saw empty shelves and assumed they needed more trucks.

The Real-Time Problem

Most supply chain systems update daily. Some update weekly. A few still rely on monthly reports.

But disruptions don't wait for batch jobs.

In 2022, when Russia invaded Ukraine, companies had hours—not days—to figure out their exposure. Did they have suppliers in the region? Did their suppliers have suppliers there? Were critical materials routed through affected areas?

Companies with context graphs could answer these questions in minutes. They could see which entities were in affected regions, trace the downstream dependencies, and start activating alternatives before the first shipment failed to arrive.

Companies without this visibility spent weeks in spreadsheet hell, sending urgent emails to procurement teams, hoping someone somewhere knew the answer.

Simulation Changes Everything

Knowing your graph is step one. Understanding the causal relationships is step two. Simulating what happens when something breaks is step three—and this is where it all pays off.

Let's say you find out that a key supplier is about to go offline for three weeks. Maintenance, labor dispute, flood—doesn't matter why. What do you do?

With a static supply chain model, you make some calls and hope for the best. With a context graph that encodes cause and effect, you can simulate.

You remove that node from the graph temporarily. You run the flow calculations: how much inventory do you have at each downstream point? When does each buffer run out? Which alternative suppliers can absorb the demand? What's the cost and lead time of switching?

The simulation gives you a ranked list of options. Maybe you can expedite from Supplier B at 15% premium. Maybe you can shift production to a different plant that uses a different component. Maybe you need to tell three customers their orders will be late—but only those three, not everyone.

This is what Procter & Gamble did when Winter Storm Uri hit Texas in February 2021. Chemical plants across the Gulf Coast shut down, affecting everything from resins to refrigerants. P&G's supply chain team ran simulations within 24 hours of the storm hitting, identified which products would be affected, and started reallocating inventory and shifting production before their competitors knew they had a problem.

From Reactive to Predictive

Most supply chain teams operate in reactive mode. Something breaks. They scramble to fix it. They add the failure to a risk register. They move on until the next fire.

This approach worked when supply chains were simpler and disruptions were rarer. It doesn't work anymore.

Between 2020 and 2023, global supply chains experienced more major disruptions than in the previous two decades combined. Pandemic lockdowns. Port congestion. Container shortages. Semiconductor scarcity. Geopolitical tensions. Climate events. Each disruption exposed companies that couldn't adapt fast enough.

A context graph shifts you from reactive to predictive. Instead of waiting for the fire, you're modeling where fires are likely to start and what happens when they do.

You can run scenarios: What if our main port of entry closes for two weeks? What if this supplier's region experiences a drought? What if tariffs increase by 25% on this route?

The graph gives you answers in hours instead of weeks. You can build contingency plans for specific scenarios. You can pre-negotiate capacity with backup suppliers. You can stockpile strategically at the right nodes.

The Path Forward

Building a context graph isn't a weekend project. It requires data integration, supplier cooperation, and ongoing maintenance. But the companies that have done it report the investment pays off within the first major disruption they face. (If you're curious how we approach this at Ergodic, see how our platform models supply chains as interconnected systems.)

Operating blind in an interconnected world isn't really an alternative at all. Without this capability, you are just waiting for the next crisis to find you before you find it.

Your supply chain is already a graph. The question is whether you can see it.