Skip to the content

DocumentationAnalytics

Operate and govern

Analytics

Incidents and response, computed from the records

IT Operations Centre → Analytics shows an organisation's incidents and its response over a period it chooses. Every figure is computed from the organisation's own records, read under row-level security, by a written-down method. The same period gives the same figures, and the fingerprint shows it.

The method analytics@1

The rules are in apps/api/src/analytics.controller.ts (METHOD, computeAnalytics). A changed rule is a new method version, never a silent change.

  • Period. From the start of the first day to the end of the last, in UTC; 30 days to today when not chosen; at most 366 days. The incidents counted are those detected in the period. Its end is the end of the last day, or now if that is earlier. Anything recorded after the end does not count, so a past period stays as it was.
  • Totals. Opened (detected in the period), acknowledged, resolved, and not resolved.
  • Time to acknowledge. From detection to the incident's first move out of New (acknowledged_at).
  • Time to resolve. From detection to its latest resolution (resolved_at, or closed_at when it was closed without one). A reopened incident counts from its latest resolution.
  • Distributions. Of incidents with a time: how many, the median, 90th and 95th percentiles (nearest rank), the mean, and how many took up to 5, 15, 30 and 60 minutes, 4 hours, 1 day, 1 week, or longer. Minutes to one decimal.
  • Over time. Opened and resolved per day, week (Monday first) or month, among the incidents of the period.
  • Breakdown. By service, category, severity (as the incident has it now) or team (the assigned team): opened, resolved, and the median times.
  • Attainment. Against the organisation's own response target for each severity. Within: done in time. Missed: done later, or not done and the target already passed at the end of the period. Pending: not done and the target not yet passed; counted neither way. The percentage is within of measured (within plus missed). Without a target there is no attainment, not a number.
  • Fingerprint. SHA-256 of the method, the period and every figure, with the ids of the incidents counted.

Response targets

Targets are the organisation's, never OpsCommand's; none ships. Per severity: acknowledge within, resolve within, or both, in minutes. Acknowledging cannot be later than resolving. A target is changed, never deleted, and every version is in the audit chain (ops.response_targets, ops.save_response_target, migration 0051_analytics).

Who, and on which plan

  • Every member reads the analytics and the targets. organisation.manage sets targets.
  • Both need the plan feature advanced_analytics (Professional and Enterprise; not Starter). The database refuses without it (ops.analytics_gate).

Export

  • GET /v1/organisations/{org}/analytics?format=csv: the incidents counted, one row each (reference, service, category, severity, team, detected, acknowledged, resolved, minutes to each), with a last line naming the method, the period and the fingerprint. Cells that start with =, +, -, @ are prefixed so a spreadsheet does not run them.
  • ?format=json: the figures, as the API returns them.

API

  • GET /v1/organisations/{org}/analytics?from=&to=&grain=day|week|month&by=service|category|severity|team&service=
  • PUT /v1/organisations/{org}/analytics/targets/{severity} with acknowledgeMinutes and resolveMinutes.

Service-level objectives (availability, latency, error rate) are measured on their own, in the IT Operations Centre (step 4.9, GET …/slo-status).