Sample SLA report

Every number in this editorial example is sample data and adds up. Its timestamps and exclusion logic follow the product record, but Perstat does not generate this report automatically.

Summary · api.example.com · Q2 2026

MetricValue
Availability, gross99.925%
Availability, after excluding announced maintenance99.994%
Confirmed incidents1
Maintenance windows1
Check interval30 s
Checks executed1,572,480 (6 regions, 91 days)
Regions confirmingna (Newark), eu (Frankfurt), as (Singapore), sa (São Paulo), af (Johannesburg), oce (Sydney)

Windows

Window (UTC)CauseDurationCounted against SLA
2026-05-03 02:41:12 to 02:48:24incident, quorum-confirmed00:07:12yes
2026-05-17 22:00:00 to 23:30:00maintenance, announced01:30:00no, excluded

The exclusion matches an announced maintenance window and carries its announcement timestamp and real boundaries.

Incident record

Time (UTC)Event
02:40:42check fails, eu (Frankfurt)
02:41:12quorum confirms, na (Newark) and as (Singapore) agree. Incident opens.
02:41:12page sent, push
02:44:03acknowledged
02:48:24recovery confirmed across regions. Incident closes.
02:55:00post-mortem opened, 2 action items

The window in the SLA table above, 02:41:12 to 02:48:24, runs from confirmation to confirmed recovery. An incident starts at the quorum confirmation, not at the first failed check or a rounded minute. The 30 seconds before confirmation are visible in the record, neither silently stretched nor shrunk. The sample monitor runs a policy of 1 failed check per region with a quorum of 2 regions, not the default of 2 consecutive checks.

Why this holds up

  • Timestamps come from the monitoring system itself, not from a human recollection.
  • Announced maintenance stays visible and is excluded by its window.
  • Later false-positive curation is a separate action with an actor, a reason, a previewed before-and-after effect, and a reversal.
  • Availability is floored, never rounded up. An SLA number should be a lower bound you can defend.
  • Linked status-page components use the same monitor and incident record. Publishing remains an explicit action.

Recompute the math yourself: the quarter’s 91 days are 131,040 minutes. Confirmed incident time of 7:12 and announced maintenance of 90:00 make 99.925% gross. Excluding the announced maintenance turns it into 99.994% counted.

This is an editorial example, not an assurance report or a regulatory filing. Each monitor view offers a PDF report for 7, 14, 30, or 90 days on every plan. The report is not signed or certified. Perstat provides an organization CSV plus manifest manually, on request.

Start monitoring free or see the product tour.