An error budget is the allowable unreliability a service can experience without breaching its service level objectives (SLOs). It turns reliability into a delivery decision: teams can release features while the budget remains, but should prioritize stability when it is exhausted. section 8.1.3
Use budget status to guide engineering priorities:
- While the service is within budget, teams can release features or make risky changes. section 8.1.3
- When the budget is exhausted, shift effort toward stability and reliability issues. section 8.1.3
- Track how quickly the budget is consumed to decide when to slow deployments, improve observability, or invest in resilience. section 8.1.3
Availability target
For an API with a 99.9% monthly availability SLO, the remaining 0.1% is the monthly error budget—about 43 minutes of downtime. section 8.1.3
Table 8.2 Some example SLOs for a service, each of which is a well-defined performance target
| SLO Type | Description | Example Target |
Availability |
Ensures that the API is reachable and functioning for consumers. |
At least 99.9% of requests succeed over a rolling 30-day period. |
Latency |
Measures how quickly the API responds to requests. |
95% of requests complete within 200 milliseconds. |
End-to-End Processing (Asynchronous) |
Measures how long it takes for published events to be processed by downstream consumers. |
99% of events are processed successfully within one minute. |
Data Freshness |
Ensures that derived views, caches, or replicas reflect recent updates. |
Replicated data is updated within 60 seconds of a change. |
Example service SLOs show the kinds of targets that define the reliability limits an error budget helps teams manage. section 8.1.3


