Error Budget
Also known as: reliability budget
An error budget is the amount of unreliability an SLO permits — the complement of the target. A 99.9% SLO over 30 days allows 43m 12s of downtime, or 0.1% of requests failing. It converts reliability from an argument into an accounting balance that both product and engineering can read.
Last reviewed · Part of the Architecture Glossary
In practice
The budget is what makes an SLO operational rather than decorative. With 1,000,000 requests in the window and a 99.9% target, the budget is 1,000 failed requests. Spend 400 on a bad deploy and 600 remain — a fact, not an opinion, and one a product manager can reason about.
The policy attached to it is the real artefact:
| Budget remaining | Policy |
|---|---|
| > 50% | Ship freely; consider taking more risk |
| 10-50% | Normal caution; reliability work gets a slot |
| < 10% | Feature freeze; only reliability and rollback work |
| Exhausted | Freeze until the window rolls, plus a postmortem |
Two things this buys. It ends the "are we allowed to deploy on Friday" debate — the budget answers it. And a consistently unspent budget is also a signal: the SLO is too loose, or the team is being more conservative than the business needs.
When it matters
Release policy, on-call escalation thresholds, and prioritising reliability work against features with an actual number instead of a vibe.
Common mistake
Publishing a budget with no consequence attached. If the freeze never happens when it hits zero, the budget is a dashboard, and the SLO is back to being decoration.
See also
- SLO (Service Level Objective)An SLO is a target value for an SLI over a window — for example, 99.9% of requests succeed over 28 days.
- Burn RateBurn rate is how fast an error budget is being consumed relative to the rate that would exactly exhaust it over the SLO window.
- Availability NinesNines is shorthand for an availability target expressed as a percentage of successful time or requests.
- SLI (Service Level Indicator)An SLI is a quantitative measure of one aspect of service behaviour, expressed as the ratio of good events to valid events — successful requests over total requests, say.