The cascade is not a single failure. It is the sequence through which a single failure becomes a catastrophic one.
What a Cascade Is
A cascade is a sequence of failures in which each failure creates the conditions that make the next failure more likely or more severe. The cascade is not the initial failure — it is the propagation of that failure through the system's dependency structure, transforming a contained problem into a systemic one. The distinguishing feature of a cascade is that the total harm produced exceeds what the initial failure could have produced in isolation: the cascade generates harm that is an emergent property of the system's structure rather than the direct consequence of any single component's failure.
Cascade propagation follows the dependency structure of the system. The component that fails first creates stress on the components that depend on it. Those components either absorb the stress — if they have sufficient margin — or fail themselves, creating stress on the components that depend on them. The cascade continues until either the stress is absorbed by a component with sufficient margin or the system reaches a new equilibrium at a significantly lower performance level than it started from.
The Speed of Cascades
One of the most important features of cascades, from a management perspective, is their speed. The cascade from initial failure to systemic failure can occur on timescales that are faster than the institutional response can operate — faster than the problem can be identified, faster than the decision to act can be made, faster than the action can be implemented. This speed asymmetry is what distinguishes the cascade from the slow motion failure: the slow motion failure gives the institution time to act; the cascade does not.
The management response to cascade risk is therefore primarily preventive rather than responsive: it must be implemented before the cascade begins rather than after it has started. The structural interventions that prevent cascade propagation — the circuit breakers that isolate failing components, the buffers that absorb stress before it propagates, the redundant systems that provide alternative pathways when primary ones fail — must be in place before the initial failure occurs, because they will be needed at a speed that makes post-failure installation impossible.
The cascade cannot be managed in real time — it moves faster than management operates. What can be managed is the system's structure: the dependencies that allow propagation, the buffers that interrupt it, and the redundancies that provide alternative pathways when the primary ones fail. Managing cascade risk is an engineering problem, not a crisis management one.
Discussion