Service-level indicators (SLIs) are measurements of service behavior, while service-level objectives (SLOs) define target levels for those measurements over a period. They help teams move from vague statements such as 'the site should be fast' to measurable reliability goals.
Choose user-centered indicators
Useful indicators may include successful request rate, page response latency, checkout completion or API availability. Select measurements that reflect what users experience. A server process being up does not prove that a user can complete the intended task.
Define an objective and window
An SLO might specify that a target percentage of eligible requests complete successfully within a latency threshold over a rolling window. Define the eligible request population, exclusions and measurement source precisely so the result cannot be interpreted differently by different teams.
Use error budgets thoughtfully
The gap between the objective and observed reliability is an error budget. It can inform whether a team should prioritize feature delivery or reliability work. It is not a reason to ignore severe incidents or security problems.
Connect alerts to action
Alert on conditions that require a human response, such as a fast burn of the error budget or a sustained user-visible failure. Avoid paging on every small fluctuation. Link alerts to dashboards and runbooks.
Checklist
- Choose indicators tied to user outcomes.
- Define measurement windows and eligible traffic.
- Set objectives that reflect business needs.
- Use error budgets to inform trade-offs.
- Make alerts actionable and link them to runbooks.