Totals and trends

GET/v1/stats/outages answers "how many" questions in one call, without paging through the archive.

curl -H "Authorization: Bearer $AOT_API_KEY" \
  "https://api.aussieoutagetracker.com.au/v1/stats/outages?group_by=network&type=unplanned"

{
  "data": [
    {
      "key": "ausgrid",
      "outages": 25,
      "customers_affected": 1205,
      "median_duration_minutes": 209
    },
    {
      "key": "ausnet",
      "outages": 14,
      "customers_affected": 4260,
      "median_duration_minutes": 211
    }
  ],
  "window": {
    "from": "2026-09-20T02:28:31.420Z",
    "to": "2026-10-20T02:28:31.420Z"
  },
  "meta": {
    "data_as_of": "2026-10-20T02:28:26.234Z"
  }
}

Each row is one group:

• key — a network id, a state, a cause category, a date (YYYY-MM-DD) or a month (YYYY-MM), depending on group_by.
• outages — how many outages in the group started in the window.
• customers_affected — the sum of customers_affected across the group's outages that carry a count. null when none does.
• median_duration_minutes — the median duration of the group's restored outages.

window echoes the window used.

Choosing the split

day and month are cut in each outage's own state's local time, so a Perth outage at 11pm on the 31st counts for that day in Perth, not the next day in Sydney.

Filters and window

from and to set the window; it defaults to the 30 days before to, and to defaults to now. A window may be up to 400 days. state, network, type and cause narrow it the same way they do on GET/v1/outages.

# Unplanned outages by cause in Queensland, this year to date
curl -H "Authorization: Bearer $AOT_API_KEY" \
  "https://api.aussieoutagetracker.com.au/v1/stats/outages?group_by=cause&state=qld&type=unplanned&from=2026-01-01T00:00:00Z"

That window starts before recording did. Nothing earlier than each network's start date exists, so there is simply nothing there to count.

Reading totals honestly

• Cancelled work is not counted. It did not happen.
• Compare like with like. Networks differ in what they publish: some publish no customer counts at all, some list planned work weeks ahead and some only once it starts. A network with no counts adds outages to a total but no customers. GET/v1/networks gives each network's publishes shares; read them before ranking networks against each other.
• The archive has a start date per network. A trend that crosses a network's start date shows a step that is the archive beginning, not outages rising. Start dates are on Networks and coverage.
• Durations are of restored outages. median_duration_minutes ignores outages still open and cancelled work. Outages explains how a restoration time is known.