Skip to content

How Uptime Is Calculated

Monitoring 11 min read
On this page

Uptime and availability are the same number under two labels: the share of the time you selected that a monitor was working. That number is the headline figure on the dashboard, on Uptime, Transaction, API and Heartbeat monitor pages, on public status pages and in scheduled reports. It is the figure that ends up in service-level agreement (SLA) conversations too. Retention policy lives in Data and Log Retention. To schedule a window, see Maintenance Windows.

The Formula #

Code
availability = (range − maintenance − downtime) ÷ (range − maintenance)

All three quantities are in seconds:

  • range: the date range you selected, end minus start. If the monitor is younger than that range, the range starts at the monitor's creation date.
  • maintenance: seconds this monitor spent inside a planned maintenance window during that range.
  • downtime: seconds this monitor spent inside an open critical incident during that range, not counting any part that fell inside a maintenance window.

Uptime, downtime and maintenance seconds always add up to the range, so the downtime figure printed beside the percentage is consistent with it.

What Counts as Downtime #

Uptimia measures downtime from incidents. A failed check on its own adds no seconds. For each critical incident that touches your range, Uptimia counts only the overlap:

  • An incident that started before your range is counted from the start of the range.
  • An incident that is still open is counted to the end of the range.
  • Overlapping maintenance comes out of that overlap before it is added.

Because incidents are the unit, everything that shifts an incident's start also shifts the number. A monitor with an alert delay and outage confirmation does not open its incident on the first bad check. Read How Uptimia Monitoring Works for what has to happen before an incident exists.

Only Critical Incidents Count as Downtime #

A degraded incident adds no downtime seconds. A speed monitor sitting amber over a web-vitals budget, a domain 5 days from expiry, a standard-tier blacklist listing: each one is listed and alerted, but the percentage does not move.

Expiry warnings are the exception among critical incidents. An SSL certificate inside its last 45 days, or a domain inside its last 3 days, opens a critical incident, but the site still works, so those seconds are not downtime. An expired or invalid certificate and an expired domain do count. See Monitor Statuses for which checks produce which severity.

An automatic Security incident is never downtime either. It records a finding, such as a malware listing or another security problem the scan found, while the site stays up, so the monitor reads Warning and its percentage does not move. See Security Monitoring.

Manual Incidents #

A manual incident counts by the severity you pick when you open it. Critical counts as downtime on every monitor type that has manual incidents (all except Real User), a Security monitor included. Degraded adds no seconds. If an automatic incident on the same monitor overlaps a critical manual incident, the shared time counts once.

Changing How Much an Incident Counts #

On Professional and above, you can override how much of an incident counts as downtime. Open the incident, then More actions → Availability impact, and pick Counts in full, Doesn't count or Count part of it (a start and end inside the incident). Doesn't count and Count part of it need a reason, which shows in incident lists. On Incidents → History you can select several incidents and click Don't count or Count in full.

The Availability impact dialog on an incident page, with Doesn't count selected, a reason typed in, and the counted downtime reading 0s of 6m 59s.Enlarge
  • Only the downtime seconds change. The incident, its alerts, its logs and its recorded duration stay as they were. The override applies before maintenance is subtracted.
  • Every figure follows it. Dashboard, monitor pages, charts, scheduled reports and public status pages all use the counted seconds. Charts can show the previous figure for up to 60 seconds.
  • Only incidents that count as downtime can be overridden. Warnings, pre-expiry SSL and domain incidents and automatic Security incidents add no seconds to begin with.
  • Each change is logged in the account activity log and can be undone. After a downgrade, existing overrides keep applying, and restoring an incident to Counts in full stays allowed.

Planned Maintenance Leaves the Calculation Entirely #

A maintenance window is neither downtime nor free uptime. Uptimia removes its seconds from both sides of the fraction, so the window can neither damage nor flatter the percentage. Uptimia merges overlapping windows before it subtracts them, so time covered by two windows counts once.

  • A check result that arrives while a monitor is in maintenance is dropped on arrival. Uptimia never logs it, opens no incident from it and sends no alert. The same is true for a paused or suspended monitor. The two inbound families are the exception. A paused or in-maintenance heartbeat still logs its pings, and a server keeps ingesting agent samples through a pause, a window or a suspension. In both cases only the alerting goes quiet.
  • A monitor that spent your whole selected range in maintenance has no measured time left, so it reads 100%.

The public status page is the exception. Its per-monitor 90-day percentage and its aggregate day-by-day availability ribbon remove maintenance seconds from the denominator only. Downtime inside a window still counts there, so a status page and a monitor page can disagree for a monitor that was down during a maintenance window. See Status Pages.

The Fleet Average on the Dashboard #

The Availability cell at the top of the dashboard is the plain average of each monitor's own percentage, not a seconds-weighted total. Every monitor counts equally. A monitor checked every 30 seconds and one checked once a day move the fleet figure by the same amount, whatever their traffic. One noisy monitor can pull the number down as far as a genuinely critical one.

Four kinds of monitor are excluded from the average entirely:

ExcludedWhy
Paused monitorsNothing is being measured. The Paused cell's own sub-line says "excluded from availability"
Suspended monitorsNothing is being measured
Real User monitorsReal User Monitoring reports visitor timings, not availability, so it has no percentage to average
Monitors awaiting dataNothing has been measured yet: a server whose agent has not reported, a heartbeat that has never received a ping, or a DNS monitor with no completed sweep

Monitors in a maintenance window stay in the average. Their own percentage already has the window removed from both sides.

The Uptimia dashboard Overview strip, with the fleet Availability percentage over its 'fleet avg · outages only · last 7 days' sub-line and the Paused count marked 'excluded from availability'.Enlarge

The cell is colored on the fleet figure: green at 99.5% and above, amber at 95% and above, red below that.

When the Availability Cell Shows an Em Dash #

If nothing in view can produce an availability figure (every monitor is paused, suspended, awaiting data or a Real User monitor), the cell shows — with the sub-line no uptime monitors rather than 0.00%. A red 0.00% would read as a total outage, which is the opposite of the truth.

Rounding #

Uptimia rounds the fleet figure to 4 decimal places, and then to 2 if it came out below 99.99. The dashboard's fleet Availability cell and a monitor's own Availability cell print 2 decimals, and so does the dashboard monitor list, except that an exact 100 prints there as 100%. The monitor rail beside a monitor page prints 1 decimal.

So 99.995% displays as 100.00% on the dashboard's fleet cell and in the dashboard monitor list. Uptime and Transaction monitor pages are the exception. When the range contained real downtime, these pages never round up to 100.00%; they print 99.99% instead, so a flat 100.00% on one of those pages means no measured downtime at all. Read the downtime seconds beside the percentage, or the incident count, when the distinction matters.

The Availability cell on an Uptimia monitor page, showing the percentage for the last 30 days with the downtime duration beneath it and the change against the previous 30 days.Enlarge

The Date Range Is Part of the Answer #

Uptimia computes every percentage over the range selected on that screen. The dashboard's range picker offers Today, Yesterday, Last 7 Days, Last 30 Days, This Month, Last Month and a custom range. Pick none and the dashboard answers for the last 7 days. Your choice sticks for the rest of the browser session.

There is no 90-day preset. For a quarter, pick a custom range, and check your plan's readable history first.

Why a Long Range Can Return a Short Answer #

Monitoring history is readable for a fixed window per plan. Free gets 7 days, Enterprise 730 days. On the screens that apply the clamp, Uptimia raises a start date older than that window to the window's edge, and the answer comes back for a shorter range than you asked for. The full ladder is in Data and Log Retention.

  • The clamp deletes nothing. The rows the clamp hides are still there; the query window is what moves.
  • Upgrading widens the window instantly, with no backfill and no waiting, but only as far back as data still exists. A nightly housekeeping pass prunes old rows on whatever plan you were on at the time. On Free that removes resolved incidents older than 60 days. Upgrading a long-standing Free account does not bring back a year of history. The full picture is in Data and Log Retention.
  • You cannot ask for "everything" by leaving the start date out. A screen asked for a range with no start date falls back to its own default window (24 hours, 7 days or 30 days, depending on the chart), and the clamp applies to that.

Why Two Screens Can Disagree #

Three mechanisms make two screens disagree, and they stack.

The dashboard does not apply the retention clamp; Uptime, Speed and Security monitor pages do. On the dashboard the fleet number covers exactly the range you picked. Your plan does not matter. An Uptime, Speed or Security monitor's own Availability cells, its charts and its history list raise the start date to the plan window first. Transaction, Heartbeat, SSL, Domain, DNS, Blacklist, API and Server pages answer the range as asked. On a Free account, a custom 90-day range gives a 90-day fleet figure on the dashboard and a 7-day figure on an uptime monitor page.

The fleet average is unweighted. It will not equal the average you get by adding up seconds across your monitors, and it will not equal any single monitor's figure.

The ranges are separate. The dashboard range picker and a monitor page's own range are set independently. Confirm both before comparing two numbers.

Against an outside tool, add two more causes. The other tool has a different check interval and a different confirmation rule, so it sees different incident start times. And it knows nothing about your maintenance windows, which Uptimia removes from the denominator.

Where the Same Number Appears #

SurfaceRange it coversClamped to plan history
Dashboard Availability cellThe dashboard range picker, default last 7 daysNo
Monitor detail Availability cell and chartsThe monitor page's range pickerUptime, Speed and Security only
Public status page timelineTrailing 90 daysNot applicable (public page); maintenance comes off the denominator only
Scheduled reportsThe report's own periodNot applicable; the report covers the period set in the report itself

The dashboard, monitor pages and scheduled reports all run the same calculation, including the maintenance exclusion. The public status page runs its own 90-day timeline code, and it removes maintenance from the denominator only. Availability impact overrides apply on all of them.

Answering an SLA Question #

Quote the monitor's own Availability figure for the exact contractual period, not the dashboard fleet average. The fleet average is a health summary across unrelated monitors and means nothing in a contract.

To produce the evidence behind it:

  1. Open the monitor and set its date range to the contractual period.
  2. Read the Availability percentage and the downtime duration in the same cell.
  3. Go to Incidents → History, set the same range, filter to the monitor, and click Export CSV for the incident list. The Availability impact, Counted seconds and Impact reason columns show any incident you set to count less than its recorded Duration.
  4. State whether maintenance windows are excluded. They are, in every Uptimia figure. If your contract counts planned maintenance as downtime, add it back from the maintenance schedule.

Wider export options are in Exporting Your Data. The report that mails this out on a schedule is in Scheduled Reports.

Was this helpful?

Your answer helps the team improve this article.

Still need help?

Uptimia Bot answers right away and passes you to the team when it can't.

Source: https://help.uptimia.com/help/a/how-uptime-percentage-is-calculated