Data
Build an SLO dashboard
Error budgets from real metrics, with the burn rate that should page.
Fill it in
What they do and what good looks like.
Where they come from.
Multi-window, so a slow burn still pages.
Your prompt
Build SLOs and an error-budget view. [the services] Metrics: [your metrics] Define the SLI from the USER's perspective. "Requests that returned 200" is not what a user experiences; "requests that returned a correct response within 500ms" is, and the difference decides whether the SLO is worth anything. Set the target from what users actually need, not from what you currently achieve. An SLO set at last quarter's number guarantees no improvement and no headroom. Alert on BURN RATE with multiple windows. A threshold alert on the error rate pages for a blip and misses a slow burn that eats the whole month's budget in three days. Show the budget in minutes remaining, not as a percentage. "22 minutes left this month" changes behaviour; "99.4% availability" does not.
Use Build an SLO dashboardOpens with everything above already filled in.
Why this works
"Requests that returned 200" is not what a user experiences. This defines the SLI from their side, alerts on burn rate across multiple windows so a slow burn still pages, and shows the budget in minutes left rather than as a percentage.