A system is healthy not when its output holds, but when it still works as one purposeful whole — when effort is not rising faster than what that effort delivers. Global trade’s output is holding. Its effort is rising. That gap is the whole story.
The natural dial is volume: total goods moved, year over year. On that dial the story is reassuring — world merchandise-trade volume grew about 4.6% in 2025, well above the ~2.4% the WTO had forecast, lifted by demand for AI-related goods. Nothing resembling collapse.
Our reading is that volume is the wrong first dial. Fragmentation does not arrive as a steady decline in how much is carried. It arrives first as rising variance in whether, when, and by which route it is carried. A system losing its integration does not ship less at first — it ships the same amount with more drama: wider swings in transit times, sharper insurance repricing, more re-routings, longer and less predictable queues. The mean holds while the spread around the mean widens.
The recent record reads this way exactly. Within a few years the system absorbed a pandemic whiplash in container demand, a war reshaping energy cargoes, attacks that emptied a primary chokepoint, and drought constraining another. In the first two months of 2024, traffic through the Suez Canal fell roughly 50% year-on-year and the Panama Canal about 32%; both ran more than 40% below their recent peaks, and Panama’s drought cut daily transits from a normal ~38 to about 24 by late 2023. Each event was absorbed; volume recovered each time. But the shape changed — and the clearest measure of it is reliability. Container schedule reliability (the share of ships arriving when promised) ran near 78% before the pandemic, fell to roughly 50–55% across 2024 (53.8% in December 2024), recovered to 65–68% through 2025, and opened 2026 at 62.4%. The width of that swing, not its level, is the signal.
The honest limit of that signal, stated plainly: schedule reliability swings hardest during shocks and recovers between them, so a wide swing alone is consistent with ordinary shock-and-recovery, not only with secular decline. The decisive test — the one this report has not yet run on itself — is whether the baseline variance between shocks is creeping up, not merely whether the spikes are large. We flag that as the first thing to check, and as a falsifier below.
The second reading concerns what the responses themselves mean. Re-routing around the Cape, friend-shoring of supply chains, “de-risking”, near-shoring, the duplication of supplier networks across political blocs — usually reported as separate cost stories, these are structurally one story: this is what it looks like when an integrated system reorganises into something less integrated, while staying busy throughout.
Integration was never just infrastructure. The efficient trade web of recent decades rested on an assumption so reliable it became invisible: that any cargo could move between almost any two points, and the political layer would not interfere. What is being withdrawn now is not capacity — ships and ports still exist — but that assumption. And the data shows trade redrawing itself around trust lines rather than cost lines: over 2017–23, trade volumes grew about 2.5% more slowly each quarter between geopolitically distant countries than between close ones, and at the peak in late 2022 trade between non-aligned blocs had fallen about 15%. Friend-shoring is not a policy fashion; it is the visible surface of a network deciding anew who is inside and who is outside — a process that can run for years while every actor behaves rationally and every quarterly report looks ordinary.
Here the reading turns uncomfortable, because the industry’s best year can be the system’s worst signal. A tonne-mile — one tonne of cargo carried one mile — is shipping’s basic unit of demand. When routes lengthen, tonne-miles rise even if not one extra tonne is traded. In 2024, vessel re-routing pushed ton-mile demand to a record +6% — almost three times the growth in trade volume — as ships diverted around the Cape of Good Hope (arrivals there surged 89%); Clarksons logged the strongest tonne-mile growth in over a decade.
Two mechanisms, not one — and they reverse differently. Honesty requires splitting that phrase. Some of the extra distance is reversible security-shock re-routing — ships going around the Cape because the Red Sea is dangerous — and it snaps back the day the strait is safe. The rest is persistent trust-withdrawal — trade redrawing itself along bloc lines, cargoes detouring around whole jurisdictions, the shadow fleet leaving the shared system — and it does not reverse when one shock ends. Only the second is genuine disintegration; the first is absorption, at cost. This report’s directional claim rests on the second, and a careful reader weighs the two apart rather than reading all the extra distance as decay.
This is why volume-watchers and coherence-watchers look at the same ocean and disagree. Volumes stable, tonne-miles up, fleets busy: by the stores, the system is sound. The same numbers read the other way say the system is working harder, over longer distances, at higher insurance cost, with more variance, to deliver the same underlying trade. Effort is rising faster than output — and that ratio, not the output line, is where erosion shows first.
We read any system on two clocks: a solvency clock (the stores — what has been built and earned) and a coherence clock (whether the system still works as one purposeful whole). They can run decades apart, and the gap between them is where masked decline hides.
Fleet capacity, ports, shipbuilding order books, the institutional architecture of trade — these stores are immense and not in question. The world fleet reached about 112,500 vessels and 2.44 billion deadweight tonnes by January 2025 (up from 2.35 billion a year earlier), and the container order book hit a historic high, with 4.4 million TEU contracted in 2024. Nothing here would trouble anyone — which is exactly why it proves little: stores measure what was earned, including what was earned under conditions that no longer hold.
Four lines, the same four we read in any institution:
The two clocks combine into the reading that matters operationally — not when something breaks, but what kind of shock this system absorbs and what kind finds the flank:
The principle in each case is the same: track variance and renewal, with thresholds agreed in advance. A threshold crossed is a fact; a mood is not.
A shipping company: the spread, not just the level — variance of transit times and schedule reliability on your trades against a fixed baseline; your own contract-versus-spot mix as a relationship-renewal indicator (are counterparties re-choosing you, or are you both just riding switching costs?); the share of revenue earned from deviation (re-routing, compliance premiums) versus carriage — the tonne-mile paradox on your own income statement.
A port: dependency concentration — share of throughput tied to a single chokepoint, lane or transshipment partner; hinterland diversity; call-pattern variance. A port can gain volume from re-routing and simultaneously become a more exposed node in a thinner network.
An insurer: the dispersion of war-risk and chokepoint premiums across routes as a live map of where the market prices withdrawn trust; the growth of the uninsured or alternatively-insured fleet as a measure of how much of the system has already left the shared trust layer; and the correlation of exposures that were priced as independent when the network was more integrated.
Be clear about the evidence. This is a structural, directional read — the same lens used on institutions and states, applied to the trade system, built entirely from public data (the figures above and their sources below). The canvas modelled is deliberately bounded: the economic, trade and geopolitical layers of the system — flows, routes, chokepoints, contracts, blocs — not the whole of world politics. Within that canvas the method’s natural mode is to run the system forward and back over time to read the trajectory — whether variance, the tonne-mile gap, the bloc divergence and the trust-layer exposure are widening or narrowing — rather than to score a single snapshot. That trajectory reading is the directional claim; it is not a dated forecast, and it has not been validated against held-out outcomes. Be clear about what stands behind it: this report is a hand read on public data — an analyst applying the lens, not the output of a model run. The forward-and-back trajectory described above is how the method would read this canvas in the engine; that engine reading has not been done for trade, so nothing here rests on a simulation and we make no engine claim for it. Treat the directional shape as the claim, not any single number.
This is a directional, structural read: fragmentation arriving as variance before volume, reorganisation visible as re-routing and friend-shoring, distance substituting for trust, the trust layer as the exposed flank. It is not a dated prediction, and it carries its own falsifiers, stated so they can be checked later:
A standing section of every report: what we looked for and did not fully find, and what this method cannot read. It is not a disclaimer; it is where the reading earns its trust.
| # | Open item |
|---|---|
| 1 | This is fragility shape, not a volume forecast. It reads where variance concentrates and which buffers it consumes — not a dated prediction of trade volume or of any single disruption. |
| 2 | One figure not yet hard-sourced. The contract-versus-spot mix (the bond-renewal indicator) is argued directionally here; the specific series is the next data point to fill before this section is final. |
| 3 | Framework-derived, not engine-run. The mechanism (shock-readiness governed by internal condition, not shock size) is derived from the project’s framework; it has not been run in the engine for this trade canvas and has beaten no out-of-sample baseline. (Consistent with the method note above: a hand read on public data, not a model output.) |
| 4 | Conduct over words. Where the report reads institutional or operator intent, it is anchored in observed routing, pricing and capacity behaviour — not in stated policy. |
| 5 | Two mechanisms under one phrase. “Distance substituting for trust” fuses reversible security-shock re-routing (Red Sea→Cape, which snaps back when the strait is safe) with persistent trust-withdrawal (bloc divergence, the shadow fleet). Only the second is structural decline; the report’s claim rests on it, and the two should be weighed apart. |
| 6 | The “trust layer” is several mechanisms, not one — insurance, sanctions compliance, payments, flag-states, classification — bundled here for brevity, with the shadow fleet as the cleanest single proxy. And that fleet may be a new parallel trust system forming, not only the old one being left. |
| 7 | “Effort evaporating” may be purpose changing. Duplicated supply chains read as waste only if efficiency is the system’s purpose; an order deliberately buying resilience or sovereignty is purchasing a different purposeful whole. The lens reads coherence against an assumed purpose and cannot yet adjudicate between competing valid purposes. |
| 8 | Bloc divergence may understate connector economies. Trade rerouted through third countries (Mexico, Vietnam, India) keeps the underlying relationship alive while changing its recorded path — so some “decoupling” is re-labelling, not severance. |
Sources. Trade volume: WTO, Global Trade Outlook and Statistics (March 2026); World Trade Statistics 2025. Chokepoints: IMF, “Red Sea Attacks Disrupt Global Trade” (March 2024); UNCTAD, “Navigating Troubled Waters”; Woodwell Climate (Panama). Schedule reliability: Sea-Intelligence (2024–2026 press releases). Bloc divergence: BIS Quarterly Review, “Deconstructing global trade: the role of geopolitical alignment” (Sept 2024); Banque de France, “Trade War and Geoeconomic Fragmentation.” Tonne-miles: UNCTAD, Review of Maritime Transport 2024 & 2025; Clarksons Research. Fleet & orderbook: UNCTAD, Review of Maritime Transport 2025; Clarksons. Trade barriers: IMF, “The High Cost of Global Economic Fragmentation”; Global Trade Alert. Shadow fleet: Atlantic Council / BRS (May 2024); S&P Global (Sept 2025). Figures are public and as-reported; periods are stated rather than smoothed. Proof of concept — consistency ≠ validation. © The Great Homecoming Project.