The Machine Nobody Sees
How the logistics networks running global commerce actually work — and why they keep failing under stress.
The battery powering your electric vehicle crossed five continents before it reached your garage. Cobalt was mined in the Democratic Republic of Congo. It was refined into battery-grade cobalt sulfate at a processing plant in Fujian, China. Lithium was extracted from brine pools in the Atacama Desert in Chile, then shipped to a cathode manufacturer in South Korea. The anode graphite came from Heilongjiang province. The cell was assembled at a gigafactory outside Seoul, packed into modules, loaded into a 40-foot container at the port of Busan, and shipped to Rotterdam — a journey of around 30 days through the Suez Canal. From Rotterdam, it went by rail to a battery pack integrator in southern Germany, which assembled it into a module destined for an EV plant in Tennessee. One more ocean leg through the Panama Canal, ground freight from Savannah to Chattanooga, and the cell was in a car.
Five continents. Six countries. Dozens of carriers, customs checkpoints, and handshake agreements. When it works, you just drive the car. The machine is invisible.
Now break it.
Houthi attacks shut down the Suez route. The ship reroutes around the Cape of Good Hope — roughly 3,500 extra nautical miles, 10 to 14 additional days, $200 to $400 per TEU in added fuel and crew costs. The container arrives in Rotterdam late. The port is already backed up — berthing delays of up to 10 days because every other rerouted vessel arrived at the same time. Meanwhile, the US has slapped sweeping tariffs on strategic Chinese goods that didn’t exist when the purchase order was signed. And the DRC — which supplies over 70% of the world’s cobalt — lifted its export ban but replaced it with strict quotas through 2027, capping the only supply the entire chain depends on.
Now zoom out.
This isn’t just a bad month; it’s the new operating environment. The numbers and insights from McKinsey and Harvard tell a clear story: the old supply chain is fracturing. With a 30% drop in trade last year, $165 billion has been rerouted to friendlier shores. It’s the clearest sign yet that ‘just-in-time’ has been replaced by ‘just-in-case.
We’re watching this ‘Great Reallocation’ unfold in real time. China’s slice of the U.S. import pie has already slid from 21% to 16%, with that business moving over to places like Vietnam, Mexico, and India. But it’s not all easy wins—the nearshoring boom in Mexico is starting to sweat a bit with the USMCA review coming up this July. This isn’t just some passing trend; it’s a total structural reset as companies scramble for ‘safe’ regional networks to outrun tariff hikes.
The logistics systems we rely on were built for a world of open trade and predictable costs, but that world has eroded — one conflict and one tariff at a time.
And here’s the uncomfortable part: most companies trying to respond to this moment — redesigning sourcing, renegotiating carrier contracts, evaluating new warehouse locations — can’t draw a clear picture of the network they’re trying to fix. The data makes the point plainly. A survey of 500 import leaders by STG Logistics found that over half would’ve diversified their supply chains much sooner if they could redo 2025. The issue wasn’t a lack of options—it was a lack of end-to-end visibility that kept them from moving fast enough.
This article breaks down how to fix that, layer by layer.
Seven layers, and what’s hidden in each
A logistics network isn’t one thing. It’s seven things stacked on top of each other. Most conversations about “supply chain resilience” go sideways because people are arguing about different layers without realizing it.
Layer 0: Geopolitics and regulation — the rules nobody controls
Think of this as the operating system for global commerce — the external constraints that dictate what every layer above it is allowed to do. That OS was once designed for efficiency. Now it’s been rewritten for strategic autonomy. Trade policy has shifted from lowering barriers to deliberately building them. Regional manufacturing corridors are being carved out. Subsidies are tied to local production. Export controls now determine not just where you can sell, but who you’re allowed to work with. And a newer layer of legal obligation means companies are now responsible for proving the provenance of their entire supply chain — labor practices, carbon footprints — several tiers back from anything they actually touch.
The foundation rewrites itself. Every design decision above it has to keep up.
Layer 1: Nodes — the dots on the map
Every logistics network is made up of locations connected by routes—factories, warehouses, distribution centers, ports, crossdocks, and foreign trade zones. What matters most is how these places are arranged in relation to each other.
A framework from Marshall Fisher, nearly three decades old, still holds: network design should match what you’re shipping. Stable products need efficient networks built at low cost. Fast-moving products need responsive networks built for speed. The gap between the model and what most companies are actually running is enormous — because most networks were never designed. They grew.
Path dependency is rarely discussed. For example, a warehouse might be in Reno because a company they bought moved it there in 2014. The European distribution center could be in the Netherlands because of a tax incentive that ended years ago. These choices are rarely reviewed—until a new tariff suddenly changes the cost equation and the old setup becomes a problem, not just an inefficiency.
Legacy patterns — still here, still fragile. In a linear network, goods travel along a single route without any backup, so if one part fails, the whole system stops. Backbone structures have a main route with smaller connections feeding into it. The Suez Canal blockage in 2021 is a good example of how problems with a backbone can disrupt global trade. Daisy chains move goods step by step from supplier to factory to warehouse to retailer. This approach is common, but if one link breaks, the entire process can stop. Ring networks, which are sometimes used in automated warehouse conveyor systems, offer closed-loop redundancy but often cost more than they benefit. These designs are not outdated; most modern networks still use them, and this is often where weaknesses appear.
The modern playbook. Hub-and-spoke is the workhorse: funnel volume through central nodes, reduce connections, capture cost through concentration. FedEx built Memphis into one of the world’s busiest cargo airports on this logic. The catch is that concentration cuts both ways — a 2026 study in Scientific Reports showed centralized networks can reduce systemic risk by 16 to 50% through strategic rewiring alone, meaning the architecture is the risk variable. Mesh connects every location directly to every other — most resilient, most expensive, and impossible to run at full scale. Most real networks end up as hybrids. They use hub-and-spoke setups for steady, high-volume routes and direct links for urgent or unpredictable needs. Star and tree structures, which are tiered and common in retail, are also part of the usual options. Companies that plan and build their hybrid networks tend to perform better than those that let them develop organically over time.
What’s currently being researched? The Physical Internet reimagines logistics as an open, shared network modeled on the digital internet — standardized containers routing to available capacity, any shipper using any carrier’s infrastructure. Potential cost reductions of 25 to 50% if it scales; the barrier is that every major carrier’s advantage depends on owning a proprietary network. Multi-tier hyperconnected networks let companies coordinate in real time across every part of the supply chain. This makes operations faster and more flexible, especially in industries like semiconductors and automotive. Cloud-based platforms like Stord turn these networks into software, so companies can scale up without investing in fixed infrastructure. SVoT (Synchronized Value of Things) networks use live IoT sensor data to reroute shipments automatically rather than waiting for a planner to act. Ant colony optimization applies algorithms from biology to routing decisions — already mainstream in last-mile delivery, being researched for larger network design.
Layer 2: Links — the lanes between nodes
Links are the transportation lanes connecting nodes — ocean, air, rail, truck, and intermodal. Most companies pick their lanes once, lock them in, and rarely revisit the choice unless something breaks.
The ocean carries over 80% of global trade by volume. When Red Sea disruptions forced rerouting in 2024, container ship calls in the region dropped roughly 75%, voyages grew by around two weeks, and Asia-Europe spot rates more than doubled on some trades. Carriers restructured entire service loops — that’s active link redesign under pressure, not disruption management. EU ETS carbon requirements now add real cost to European shipping lanes on top of freight rates.
Air shipping costs four to ten times more per kilogram than ocean shipping on major routes. But for high-value goods, when you include the cost of holding inventory for an extra 30 days, air can actually be cheaper overall. That’s why it’s not as simple as ‘ocean is cheap, air is expensive.’
Rail is the underused middle option. On lanes over 700 miles, intermodal shipping — rail linehaul plus truck at each end — delivers meaningfully lower cost than road-only freight with significantly lower emissions. The catch is added handoffs and aging infrastructure. Trucking remains the most flexible option, but it has its own problems, such as a shortage of drivers, fluctuating fuel prices, and increasingly stringent regulations.
The real challenge isn’t just choosing the cheapest shipping method. A 2024 paper in the American Economic Reviewshowed that the costs of changing supply chains due to tariffs go far beyond the tariffs themselves—searching for new partners, renegotiating deals, and losing established relationships add up and are often missed in standard cost models. Most transportation management systems only optimize the routes already set up. They won’t alert you if a route no longer makes sense because the total cost has doubled. To spot that, someone needs to treat these links as design choices, not just fixed facts.
Layer 3: Flows — the physics of inventory and time
Nodes are where goods stay. Links are how they move. Flows are the rules that decide what moves, when, and how much.
The decoupling point — first formalized in 1992 — is where the network switches from push (build to forecast) to pull (build to demand). Where you place it determines your entire inventory strategy. Safety stock is the buffer against variability in demand and lead time. A landmark 1997 study showed that order variance amplifies as it moves upstream — the bullwhip effect. Rational behavior at each node creates irrational volatility across the network. Most companies can tell you their average lead time. Almost none can tell you their lead time standard deviation by lane, by mode, by season. Without that, the safety stock calculation is a guess with math around it.
Cost-to-serve is the key measure that brings everything together—and it’s often the most hidden number in logistics. What does it really cost to deliver a certain product to a specific customer using a particular route? Not the average, but the full, actual cost for that order. Most companies don’t do this analysis. They might not realize that Customer A in Munich is profitable through the Rotterdam distribution center but not through Milan, just because of the last-mile carrier. Seeing the true cost-to-serve changes decisions. Without it, you’re trying to optimize a network you can’t fully see.
Layer 4: The data layer — the source of truth that’s almost always broken
Every layer above generates data. The problem isn’t generation. It’s integration. There’s a solid argument from 2000 that visibility and velocity are the twin foundations of supply chain agility. That was written a quarter century ago. The gap is worse now than most executives realize.
A typical mid-size manufacturer runs an ERP, a TMS, a WMS, and possibly an OMS — each purchased at different times from different vendors, running on different data models. Getting a real-time, end-to-end view of a single order requires stitching together four or more systems that were never designed to talk to each other. And no single team owns the canonical definition of an order or an inventory position. Better integration without clear data ownership just means propagating bad data faster.
The numbers confirm how broken this is: McKinsey’s 2025 Supply Chain Risk Pulse found 95% visibility into tier-one suppliers, dropping to 42% at tier two. Beyond that, visibility nearly disappears. And 2025 trade regulations increasingly require proof of origin several tiers upstream — turning the visibility gap into a compliance gap.
When the data layer is fragmented, every optimization above it is local at best and wrong at worst.
Layer 5: The decision layer — the strategy no one owns
This is the layer that never makes it onto a presentation slide: who makes which decisions, when, and using what information.
In most companies, logistics decisions are split across procurement, operations, finance, and a supply chain function that’s supposed to coordinate all of them but rarely has authority to override any of them. Network design — where nodes go, how links get configured, how flows get governed — shapes cost structure, service capability, and competitive position for years. But in most organizations, it gets treated as an operations project.
That process worked when the world was stable. The “Triple-A Supply Chain” framework identified this two decades ago: the best supply chains need agility, adaptability, and alignment — not just efficiency. Recent research shows that tariff uncertainty specifically — not just tariff levels — pushes firms away from the long-term relational contracts that make networks stable. McKinsey’s Supply Chain Risk Pulse found 82% of firms say tariffs are actively affecting their networks, and most responses have been tactical. Structural redesign keeps getting deferred.
Procurement owns the suppliers. Operations owns the execution. Finance owns the budget. Nobody owns the architecture. And when the environment shifts, there’s no one in the room who can say: here’s what our network looks like, here’s where it’s exposed, and here’s what we need to change.
Layer 6: The orchestration layer — when supply chains compete against each other
At this level, companies aren’t just competing with each other—they’re competing supply chain against supply chain. Two rivals might use the same third-party warehouse, the same return truck, or the same cross-dock. Programs that pool capacity and cut down on empty trips have saved 20 to 25% in costs—savings no single company could get on its own. But this means sharing sensitive data with partners who might also be competitors in some areas. Projects like Catena-X in the auto industry are creating data spaces where companies can share supply chain info while still controlling what they reveal and to whom. Layer 6 only works if the layers below are solid. Most companies aren’t there yet, but the best supply chains are moving in this direction.
What’s breaking — layer by layer
A 2025 paper in the Review of Economic Studies confirmed what many in the field already sense: supply chains are naturally fragile. Even small disruptions can cause sudden, widespread failures. What’s happening now is proving this theory true.
The nodes are overwhelmed. When tariffs hit in April 2025, 85.6% of US importers rushed to ship goods early. Warehouses quickly filled up. Drayage costs soared because everyone needed trucks at once. This wasn’t resilience—it was a rush that showed which companies had built in extra capacity and which hadn’t. Most hadn’t.
The links are repricing. McKinsey’s Supply Chain Risk Pulse found 82% of companies affected by tariff and trade policy changes. Asia-Europe rates remained 25% or more above pre-crisis levels for extended periods. For high-tariff, China-concentrated product categories, total cost-to-land effectively doubled within 18 months. By mid-2025, 75% of goods firms reported margin declines despite raising prices. The links didn’t just get disrupted — they repriced permanently.
The flows are being made up as they go. In 2025, Foreign Trade Zones became a popular way to avoid tariffs—a clever financial move, but not a real redesign. The key points in the network didn’t change. Safety stock calculations stayed the same. Companies just added a quick fix to a system they still don’t fully understand.
The data layer can’t keep up with the new reality. In 2025, 79% of companies shifted sourcing away from China. New suppliers mean new lead times, new variability profiles, new customs requirements — none captured in the historical data planning systems rely on. The companies that had built real multi-tier visibility before the disruption hit had a meaningful advantage. Everyone else was flying blind during the most volatile logistics period in a generation.
No one owns the decision layer. 40% of US companies plan to move supply chains to North America by 2026. But CSIS found that most foreign investment in Mexico is just reinvestment, not new projects—so plans are ahead of actual spending. A World Bank study showed that much of what looks like friendshoring is really just rerouting trade through countries like Vietnam and Mexico, while still relying on Chinese factories. The network didn’t really change—it just added another stop. Who is figuring out what happens if USMCA renegotiation changes the rules that make Mexico attractive? In most companies, no one.
Before you redesign — see the machine first
The same pattern keeps showing up: companies make isolated decisions about a problem that affects the whole system. An ECB working paper measured the cost—short-term losses from breaking up supply chains are five times higher than the long-term costs. The companies losing the most right now aren’t the ones that stayed put, but those that moved quickly without understanding the whole network they were changing.
There’s a book from 2005 — The Resilient Enterprise — that argues resilience has to be designed into a network’s structure before crises hit, not bolted on as a response. Two decades and several global disruptions later, most companies still haven’t internalized the lesson.
So the first step isn’t a new strategy or a new consulting engagement. It’s visibility. Once you can actually see the machine, the path forward gets clearer fast.
The companies that held up best through the 2025 tariff shock weren’t running better networks on paper — they were running scenarios constantly, not once a year. Only 19% of supply chain leaders do this, meaning most update their strategy after the disruption hits, not before. They also knew their actual cost per lane—not blended averages—so when a route blew up, they could act rather than guess. And someone on their team owned the network as a whole. Procurement owns suppliers. Operations owned execution. But somebody owned the architecture.
KPMG’s 2026 Global Trade Outlook puts it simply: unpredictable tariffs are now a permanent part of the landscape. In a world of rolling disruptions, invisibility isn't efficiency. It's the risk.
When we started writing this, we expected to find complexity. What we found was something worse: invisibility. Companies aren’t failing because they can’t handle complexity. They’re failing because they can’t see the system they’re trying to fix.
The companies that see their networks clearly will redesign them. The ones that don’t will keep reacting to each disruption as if it’s the first one.
It won’t be the last.
This is the first article in a series about the structure of global logistics and supply chain management. Written by The Industrial Brain team. We publish analysis covering the industrial space and the forces shaping it.
Further Reading
McKinsey: Geopolitics and the Geometry of Global Trade — 2026 Update — The fullest picture of how trade flows are actually shifting: US-China down 30%, $165B rerouted, and who’s absorbing it.
STG Logistics: 2026 Supply Chain Reset Survey — 500 import decision-makers on what they actually did when the tariffs hit, and what they wish they’d done sooner.
Acemoglu, D. & Tahbaz-Salehi, A. (2025). “The Macroeconomics of Supply Chain Disruptions.” Review of Economic Studies — The mathematical proof that small shocks produce cascading, nonlinear failures. If you want to understand why the current environment breaks networks the way it does, start here.
Alfaro, L. & Chor, D. (2025). “An Anatomy of the Great Reallocation.” NBER — The most rigorous empirical account of where supply chain geography actually moved, and how fast.
Fisher, M.L. (1997). “What Is the Right Supply Chain for Your Product?” Harvard Business Review — Nearly 30 years old and still the right starting point for any network design conversation.
Sheffi, Y. (2005). The Resilient Enterprise. MIT Press — The argument that resilience has to be built in before a crisis, not improvised during one. More relevant now than when it was written.