What is Blast Radius?

Blast radius is a function of coupling. A component with many dependents and no isolation has a large radius because its failure propagates directly to everything downstream. The same component behind circuit breakers, timeouts, and fallbacks has a much smaller one because dependents degrade instead of failing.

Architecture sets the ceiling. Shared databases, shared caches, and shared control planes create wide radii because a single failure touches everything using them. Partitioning, cell-based architectures, and per tenant isolation reduce the radius by ensuring a failure is contained to one partition.

It converts a vague sense of risk into something you can reason about before acting.

• Change safety: a deploy that can affect all users is a different risk than one affecting five percent.
• Severity assessment: scope determines how an incident is classified and staffed.
• Architecture review: components with outsized radius are the priority for isolation work.
• Automation bounds: automated actions are constrained by how much they are allowed to affect.

The dangerous case is the radius nobody mapped. Undocumented dependencies, shared infrastructure that looks isolated, and implicit coupling through a common library or a shared quota all produce failures that spread further than the architecture diagram suggests.

Radius also changes with load. A dependency that degrades gracefully under normal traffic may become a hard failure at peak, when timeouts and retries interact to consume shared capacity. Measuring radius during quiet periods understates it.

Severity is a judgment about how bad an incident is, combining scope, duration, and business impact into a single classification that drives response.

Blast radius is one input to that judgment and is a structural property rather than an event property. It exists before anything fails. You can reduce blast radius through architecture in advance, whereas severity is assessed only once an incident is underway. Reducing radius is preventive work, assessing severity is response work.