Error budget

An error budget is the amount of unreliability a service is allowed to have over a period, calculated as 100% minus the service level objective target, expressed as a quantity of allowed failures, downtime minutes, or failed requests rather than as a percentage alone. If an SLO is 99.9% availability over 30 days, the error budget is the roughly 43 minutes of allowed downtime in that window.

The concept differs from simply tracking uptime because it turns reliability into a spendable resource: as long as budget remains, a team can take risks such as shipping new features or running experiments; once it is exhausted, the team is expected to pause new releases and prioritize stability work until the budget recovers. This links error budgets directly to release policy in a way a raw service level indicator number does not on its own.

Error budgets matter because they replace an open-ended argument between reliability and feature velocity with an agreed, numeric threshold that triggers a specific response. A common pitfall is defining the policy, freezing releases at zero budget, without actually enforcing it, which makes the SLO symbolic rather than a real operating constraint.