The case for the lean, just-in-time, single-sourced supply chain was settled twenty years ago and reopened five years later. The pandemic, the Suez and then Red Sea disruptions, and the 2021–22 semiconductor shortage each, on their own, would have been a generational shock. Together, they produced a permanent shift in how serious operating businesses think about supply-chain design. The shift is not back to inventory hoarding. It is toward something closer to what Nassim Taleb called antifragility: systems that, within limits, get better when they are stressed.
Why efficiency stopped being enough
Lean supply chains, optimised for the steady-state, are remarkably efficient — and remarkably costly when the steady-state breaks. The cost shows up not only in lost sales during the disruption, but in the months of operational chaos that follow as the supply chain re-tunes. The companies that came out of 2020–24 best were not those with the largest buffer stocks. They were those whose supply chains had multiple viable configurations they could shift between, with little notice and reasonable cost.
Calling this "resilience" understates it. Resilience is the property of bouncing back; antifragility is the property of having designed a system that benefits from disruption — by learning which suppliers, lanes, and inventory policies work under different stress conditions, and by becoming better-tuned each time. The supply chains that look antifragile today are not the most expensive ones. They are the most thoughtfully designed.
The new design matrix
The interesting position in this matrix is not the top-right corner. It is the path from the bottom-right — where most organizations sit today — to the top-right, without passing through expensive insurance on the way. That trajectory is achievable, and it is what serious supply-chain teams are now executing.
Four design principles that travel
Across the supply chains we have helped redesign in the last three years, four principles recur.
Multi-source by design, not by accident. The companies that came through the chip shortage best did not have two suppliers because they had been forced into a backup arrangement. They had two qualified suppliers, both running steady-state volume, with the operational discipline to shift split-ratios within a quarter. The cost of running this design is modestly higher in normal times; the cost of not running it can be catastrophic.
Modular geography. The pre-2020 supply chain optimised for the lowest unit cost globally. The post-2020 supply chain optimises for a portfolio of regional configurations, each capable of supplying its primary market with limited reliance on the others. The total cost is 4–8% higher; the variance in delivered cost is far lower.
Instrumented decision-making. The disruptions of 2020–22 surfaced an uncomfortable truth: most large supply chains could not answer basic questions ("how many days of stock do we have, by SKU, by region?") in less than 72 hours. The antifragile generation has invested heavily in supply-chain control towers — not because the dashboards are intrinsically valuable but because decisions made on stale data, under pressure, are bad decisions.
Inventory policy by SKU criticality, not by category. The cost of safety stock varies enormously by SKU. Treating an entire category as "high-priority" or "low-priority" loses information. The companies that have done this well now run dynamic, SKU-level stock policies driven by criticality, lead-time variance, and substitutability — and they hold roughly the same total working capital they did before, allocated very differently.
The supply chains that came out of 2020–24 best were not those with the largest buffer stocks. They were those that had multiple viable configurations and could shift between them within a quarter.
The case for instrumentation
The gap between the top and bottom of this exhibit — eight hours versus 110 — is the difference between catching a disruption in the first wave and arriving at the conversation after the consequences have compounded. The technology to operate at the top of the table is not, in 2026, exotic; it is a standard control-tower stack costing 1–3% of supply-chain operating cost annually. The return is asymmetric.
- Design for two configurations, not one. Multi-source as a steady-state, not as a contingency arrangement.
- Instrument before you optimise. Optimising on the wrong data is worse than not optimising.
- Hold the same working capital, allocated by criticality. The total is rarely the problem; the allocation usually is.
The next shock
The most useful question for a board reviewing its supply-chain strategy is not "are we more resilient than we were in 2020?" — almost everyone is. It is "are we still optimised for the shocks of 2020, or are we designed for the shocks we expect next?" The answer for many organizations is honestly the first. The mitigations went into the supply chain that broke; the next supply chain that breaks will look different.
Geopolitical fracturing, regional carbon policy, AI-driven demand-pattern shifts, and the slow restructuring of global shipping routes are all live candidates. Designing for them is not a forecasting problem; it is an optionality problem. The companies that build optionality into their supply-chain architecture — by region, by supplier, by transport mode, by inventory location — will not predict the next shock. They will be ready for several plausible ones.