Uptime is measured from real API traffic, request by request. A request counts against uptime when it fails with a server-side error — HTTP 5xx responses. Client errors (e.g. 400 / 403 / 404 / 429), connections closed by the client before a response was delivered (HTTP 499), and windows with no traffic are not counted against uptime. This matches common industry practice, where error rates are defined over server-side (5xx) failures only.

uptime % = successful requests ÷ (successful + server-error requests)

The 24-hour and 90-day figures aggregate these requests; the 90-day and monthly numbers are weighted by traffic. Scheduled maintenance windows may be excluded.

Each window is graded by its uptime: Operational at 99.5% or higher, Elevated Error Rates between 95% and 99.5%, Partial Outage between 80% and 95%, and Major Outage below 80%. Windows with no traffic are shown as No data.