The efficiency gains of the last 30 years were real. Just-in-time manufacturing reduced inventory costs. Shared financial clearing infrastructure reduced settlement overhead. Consolidated cloud computing reduced IT capital expenditure. GPS-based timing replaced expensive dedicated clock systems across banking, telecoms, power grids and transport. Every one of these changes made individual organisations leaner. Collectively, they made the entire system more fragile.
The fragility is not in any single component. It is in the coupling between components. When 15 Danish banks share one infrastructure layer and that layer fails, it is not 15 separate incidents. It is one combined failure that nobody measured as a combined risk because each bank was measuring its own exposure, not the shared one.
The pattern repeats because the underlying structure is the same everywhere.
The Layers Nobody Monitors Together
Three shared dependency layers underpin every industry in this series. All three are monitored inside individual organisations. None are monitored as the combined risk they represent.
Layer 1 – Time
GPS is the timing reference for financial markets, power grid synchronisation, telecoms networks and aviation. It is not a navigation system that happens to be used for timing. It is a timing infrastructure that happens to also provide navigation. When it degrades, the timestamp drift hits financial settlement, grid frequency management, mobile network handoff and aircraft separation simultaneously – because all of them are reading from the same source. Each sector monitors its own timing. Nobody monitors the shared source.
Layer 2 – Software
The CrowdStrike incident was not unusual. It was a demonstration of a structure that exists everywhere. Banks, airlines, hospitals and broadcasters all running the same endpoint security layer. One update. One failure. A global cascade across every organisation that had independently and rationally chosen the same vendor. The individual risk assessment was sound. The combined picture was never calculated.
Layer 3 – Logistics
Global supply chains consolidated around a small number of major ports, shipping lanes and carriers over 30 years of optimisation. The efficiency was measurable. The fragility was not – until the Ever Given blocked the Suez Canal for six days and held up $9.6 billion of trade per day. One vessel in one channel. Every manufacturer, retailer and hospital system that had optimised away its buffer inventory discovered simultaneously that the optimisation had transferred the buffer cost into a catastrophic risk.
Why Risk Models Miss This
The measurement problem
Every organisation measures the risk it can see from inside its own boundary. Its own systems. Its own supplier relationships. Its own regulatory exposure. What it cannot measure from inside its own boundary is the risk in the shared infrastructure it depends on, because that infrastructure does not show up in its risk register. It shows up as an assumption. The assumption that GPS will work. The assumption that the clearing network will clear. The assumption that the shared software layer is stable. Those assumptions are not risk-managed. They are inherited.
The coupling problem
The efficiency gains created coupling. Coupling creates correlation. Correlated systems fail together, not independently. Standard risk models assume approximate independence between failure events. When 15 banks share one infrastructure layer, their failure events are not independent. When the entire aviation industry shares one GPS constellation, its failure events are not independent. The model is wrong before the incident happens, because the assumptions it was built on are wrong.
The Barclays mainframe was degrading for hours while every dashboard showed green – because the dashboards were monitoring the bank’s own layers, not the combined picture across all of them.
The boundary problem
A bank’s operational resilience team monitors the bank’s systems. An airline’s operations centre monitors the airline’s systems. Neither monitors the shared timing infrastructure that both depend on. The combined picture had no owner.
The Structural Argument
The incidents described in every post-mortem of the last decade are not the result of individual failures. They are the result of shared dependencies that were never modelled as shared risks. The organisations involved did not make bad decisions. They made rational decisions inside a risk framework that did not extend to the layers underneath their own boundary. The framework was the problem.
The OCTA model extends the measurement boundary. It maps the dependency structure from incoming signal to critical operational function, across layers, across the organisation’s own boundary when the dependency crosses it. The Dependency Cascade Index measures how tightly coupled an organisation’s critical functions are to each layer in the stack. The more coupled, the faster the cascade when that layer degrades. Making that coupling visible, before the failure, is the structural argument for predictive modelling.
The working paper on operational continuity in GNSS-denied environments is available here: jameslarsson.pro/research-octa
