Buffer stock was never just inventory

On just-in-time supply chains and the resilience that left the balance sheet

The Guild of Dyers has one supplier in Sto Lat for the grade of indigo that the Ankh-Morpork textile trade requires. It has one supplier because one supplier is cheaper than two, and because holding the relationship with two suppliers requires time and administrative capacity that can be directed elsewhere when you only need one. The warehouse in Hen and Chickens Field holds four days of indigo. It holds four days because four days is adequate under normal conditions, and holding six weeks would require capital that earns a better return in a counting house than it does sitting in a dyestuff warehouse.

When the Sto Plains road closes, ice in winter, flooding in spring, bandits occasionally, it varies, the warehouse runs dry in four days. Production stops at three workshops. The tailors commissioned by two noble houses cannot complete their work. The noble houses withhold payment on grounds that the work is not delivered. The tailors cannot pay their workers this week. Their workers spend less at the market on Saturday. The market stalls reduce their orders from the wholesale merchants on Monday. The indigo supplier in Sto Lat is unaffected. The road closure was four days. The damage has travelled through seven layers of economic interdependence that had nothing to do with the road.

No single firm in this chain made a bad decision. The Guild of Dyers was rational to reduce its buffer stock. The tailors were rational to operate on commission without reserve capital. The workers were rational to spend what they earned. Each individual decision was economically correct given the information and incentives available to that firm or household. The aggregate fragility was not visible to any of them, because it was a property of the system, not of any component within it. The system was optimised for cost. The optimisation was efficient. It was also brittle, and the brittleness was invisible until the road closed.

Where the cost went

The dominant account of just-in-time logistics is that it makes supply chains more efficient by eliminating waste. Inventory held in warehouses is capital tied up unproductively. Reducing inventory frees that capital for more productive uses. Coordinating tightly with suppliers so that components arrive exactly when needed reduces storage costs, reduces the administrative burden of managing stock levels, and allows firms to respond more quickly to changes in demand by not being committed to large volumes of any single input. These are genuine efficiencies. The capital freed from warehouse inventory earns returns elsewhere. The cost savings are real.

What the dominant account does not address is where the cost of having no buffer stock went. It did not disappear. It was transferred.

Buffer stock is resilience materialised on a balance sheet. A warehouse holding six weeks of indigo is not just a warehouse holding six weeks of indigo. It is six weeks of operational continuity that will survive any disruption shorter than six weeks without requiring any management decision or emergency action. When the warehouse is reduced to four days, four days of resilience leaves the balance sheet. The capital is freed. The resilience is gone. The cost of the disruption that the buffer would have absorbed has been transferred from the balance sheet, where it was visible and priced, to the operating environment, where it is invisible until a disruption arrives and reveals it.

This transfer is the mechanism. Supply chains that appear more efficient under normal conditions have transferred their resilience costs to operating environments where those costs are not measured, not managed, not hedged against, and not visible in any accounting until a disruption makes them manifest. The cost did not go away. It moved somewhere it could not be seen.

The Volcano

On 14 April 2010, the Eyjafjallajökull volcano in Iceland began erupting in a way that produced an ash cloud over European airspace. By the following morning, most of European airspace was closed. It stayed closed for six days.

The immediate commercial impact was in aviation: airlines, airports, passengers. The structural lesson was in the manufacturing and logistics chains that had come, quietly and over two decades, to depend on air freight for components that were too time-sensitive, too high-value, or too specialised to travel by sea.

BMW, Volkswagen, and Peugeot halted production lines within seventy-two hours as electronic components sourced from Japanese suppliers and air-freighted to European factories failed to arrive. The components were not large or heavy; they could not wait for sea freight because production lines would stop before a sea shipment could arrive. Pharmaceutical companies lost active ingredients with 48-hour shelf lives, airfreighted because they could not safely be stored longer. The cut flower industry watched perishable stock valued at tens of millions of euros die on African runways and in Dutch cold stores: flowers grown for the European market in Kenya, Ethiopia, and Colombia, air-freighted to the Netherlands for auction, with zero tolerance for delay because the product expires in hours.

None of these supply chains had failed. No supplier had gone bankrupt. No factory had burned down. A volcano had erupted in Iceland, and the consequences were visible within three days in automotive plants in Bavaria, pharmaceutical warehouses in Belgium, and flower auctions in Aalsmeer. The supply chains had been operating correctly under normal conditions. The disruption had revealed costs that normal conditions had kept invisible.

The disruption lasted six days and generated analysis, conferences, and industry reports about supply chain resilience that identified single-source dependencies, inadequate buffer stock, and over-reliance on air freight for time-sensitive components as structural vulnerabilities. The analysis was accurate. Most of the structural vulnerabilities it identified were not addressed.

The River

The Rhine carries approximately eighty per cent of German inland waterway freight. In the summer and autumn of 2022, a sustained drought reduced it to near-impassable levels for nearly five months. This was not unprecedented in European climate history, though its severity and duration were unusual. It was, in the abstract, the kind of event that supply chain risk models should model.

BASF’s chemical complex at Ludwigshafen is the largest integrated chemical site in the world. It was built on the Rhine specifically because the river provides the water access and logistics connectivity that a complex of its scale requires. When the Rhine ran low, BASF could neither receive raw materials by barge nor ship finished product. Production was cut. The ripple moved outward: German power stations that received coal by river found their deliveries slowing. Diesel and heating oil shipments to southern Germany, which travel by river barge, tightened, pushing up spot prices in landlocked regions. Downstream manufacturers that used BASF products as inputs faced supply interruptions that were not caused by anything in their direct supply chains.

The total estimated economic cost of the Rhine low water episode was approximately five billion euros. A drought had cascaded through water transport, through chemical production, through power generation, through fuel distribution, through downstream manufacturing, because each of those sectors had, rationally and individually, optimised its supply chain around the assumption that the Rhine would carry barge traffic. No single firm’s decision was wrong. The system’s aggregate assumption about river navigability had no buffer for the five-month drought that everyone knew, in the abstract, was climatically possible.

The risk had not been eliminated from the supply chain. It had been transferred to the operating environment in the form of a five-billion- euro exposure that nobody held on a balance sheet.

The gas that comes from fertiliser

In the summer of 2018, European brewers faced a CO2 shortage that most consumers had no reason to expect and most brewers had no contingency plan for. CO2 is essential for carbonating beer, for maintaining pressure in kegs, and for packaging food products in modified atmosphere. It does not sound like a critical industrial input. It is.

CO2 for food and beverage use is primarily a by-product of fertiliser manufacturing. The same industrial process that produces ammonia for agricultural use produces CO2 as a side stream that is captured and sold. Several major European CO2 production facilities run on a seasonal maintenance cycle aligned with fertiliser production schedules. When multiple facilities underwent maintenance shutdowns simultaneously in the summer of 2018, CO2 supply fell sharply across multiple European markets at the same time.

Brewers had no buffer. CO2 is perishable in the sense that maintaining it in liquid form under pressure requires continuous refrigeration and storage infrastructure that most breweries size for throughput, not reserve. Belgian brewers, whose product is among the most economically and culturally significant in Europe, began rationing within days. German and Dutch producers with slightly larger operational buffers were affected within two weeks. Production was reduced. Exports were delayed. In Belgium particularly, where the brewing industry carries a political weight that would be recognisable to any observer of Ankh-Morpork, the shortage became a minor political event.

The lesson, that CO2 supply was a single-source dependency concentrated in a production process with a predictable seasonal maintenance cycle, with no buffer in the distribution chain and no contingency in most brewers’ operational plans, was documented, presented at industry conferences, and largely not addressed structurally. The UK variant of the same shortage, when CF Fertilisers closed its two domestic production sites in 2021 following a gas price spike, produced the same cascade through the same structural absence of buffer, three years after the European shortage had identified the vulnerability.

The system had been optimised around the assumption that CO2 would be continuously available. The assumption had been correct for long enough that no one had maintained the resilience infrastructure that would make the assumption optional.

Individual rationality, collective fragility

The pattern across these cases is consistent: at every individual node in the supply chain, the decision to reduce buffer stock and single-source supply was rational. The capital costs of holding inventory are real. The administrative costs of managing multiple supplier relationships are real. The efficiency gains from tight coordination with a preferred supplier are real. Any firm that chose to hold six weeks of buffer stock while its competitors held four days was carrying a cost disadvantage under normal conditions. The individual incentive to optimise is clear and powerful.

The aggregate consequence of all firms in a chain making the same rational individual decision is a system with no resilience at any node. When a disruption arrives, it does not find one node with buffer and twelve without. It finds thirteen nodes without buffer, and the cascade runs through all of them simultaneously.

This is not a failure of individual decision-making. It is a failure of the information environment in which individual decisions are made. Each firm optimises against its own costs and the behaviour of its direct counterparties. No firm in the chain has full visibility of the aggregate resilience profile of the system it participates in. No firm is paid to maintain buffer stock for the benefit of firms downstream from it. The firm that holds six weeks of indigo is paying for the resilience of the entire dye-to-garment supply chain, receiving the cost of that resilience on its own balance sheet, and receiving none of the benefit, because the benefit, production continuity for tailors two steps downstream, does not accrue to the dye warehouse.

The resilience that buffer stock provides is a public good within a supply chain. Individual firms cannot capture the full return on providing it. The rational response is to free the capital. The system consequence is fragility that appears efficient until it isn’t.

What resilience actually costs

There are three places where supply chain resilience can live, and each has a different cost structure and visibility.

The first is buffer stock: inventory held beyond immediate need. This cost is fully visible on balance sheets, accrues to individual firms, is optimised away under competitive pressure, and disappears in the JIT transition.

The second is supplier diversity: maintaining relationships with multiple suppliers for the same input, accepting higher unit costs or administrative overhead in exchange for optionality when the primary supplier fails. This cost is partially visible, is routinely sacrificed for the lower unit cost of the single preferred supplier, and disappears as procurement is professionalised and consolidated.

The third is geographic proximity: maintaining supply sources close enough to the point of use that short-cycle disruptions can be absorbed by switching to local supply. This cost is visible in higher unit prices for locally produced inputs compared to the global minimum. It is routinely sacrificed for the lower cost of global sourcing and disappears as supply chains extend.

When all three forms of resilience are eliminated simultaneously, as they were across many European supply chains through two decades of optimisation, the system’s resilience is entirely dependent on the continued normal operation of every upstream input, every logistics route, and every supplier at every tier. The assumption of normal operation was correct for long enough to validate the optimisation. It became incorrect when a volcano erupted, or a river ran dry, or a maintenance schedule coincided across facilities, or a pandemic disrupted production patterns globally.

None of these events were unforeseeable in the abstract. All of them were events that supply chain risk models could have modelled and that buffer stock, supplier diversity, or geographic proximity would have partially absorbed. The costs of providing that absorption had been moved off the balance sheet. They returned, at scale, as operating losses when the disruption arrived.

The Dye Warehouse and the Counting House

The capital freed from the Hen and Chickens Field dye warehouse earns a return in a counting house somewhere. That return is real. It appears in accounts as productive investment. The resilience it used to provide does not appear anywhere until the road to Sto Lat closes and the workshops stop.

When the workshops stop, the cost is attributed to the road closure. The road closed; production stopped; the road is the cause. This attribution is accurate in the sense that no road closure means no disruption. It is incomplete in the sense that a road closure of four days would not have stopped production if the dye warehouse had held six weeks of indigo. The fragility that converted a four-day road closure into a seven-layer cascade was in the supply chain before the road closed. It was created, over years of rational individual optimisation, and it was invisible because it only manifests in the presence of a disruption.

The cost of eliminating buffer stock is paid at the warehouse, visibly, every year, as freed capital that earns its return. The cost of not having buffer stock is paid across the supply chain, invisibly, in the year when something goes wrong. The two costs do not appear on the same balance sheet at the same time. The first cost is managed. The second cost is absorbed.

The Guild of Dyers will rebuild its indigo stock when the road reopens. By the following season, the warehouse will be back to four days, because four days is adequate under normal conditions, and holding six weeks would require capital that earns a better return elsewhere.

The road will close again.