Error budget

Error budget is the allowed unreliability implied by an SLO. If availability SLO is 99.9% monthly, budget is ~43 minutes of bad events per month. Teams spend budget on launches, experiments, and accepted risk—or stop shipping when budget is exhausted.

					remaining budget = allowed bad events - actual bad events
				

| Budget state | Typical response | |--------------|------------------| | Healthy | Ship features, tolerate some risk | | Burning fast | Freeze risky changes, fix reliability | | Exhausted | Reliability work only until recovery |

Product and engineering share the budget conversation—it is not only ops. Burn-rate alerts give early warning before users notice SLA risk.

On interviews: balancing feature freeze vs innovation, partial outages consuming budget, and multi-service budgets.

Common pitfalls: SLO without enforcement; ignoring budget during crunch; measuring budget on the wrong SLI.

The trade-off is speed of delivery versus protecting user trust during change.

Checklist:

• Compute budget from explicit SLO.
• Track burn visually for teams.
• Define policy when budget is low.
• Tie postmortems to budget impact.