How it works
This is how we turn user reports into the status of a service.
- 15minutes
- Length of each interval
- 21days
- History for the baseline
- 2,5×
- Alert threshold
- 5×
- Incident threshold
From one person's click to the service status
-
1
Report
A person tells us what problem they are having.
-
2
Interval
Reports are added up in blocks of 15 minutes.
-
3
Comparison
Each block is compared with the usual level for that time of day.
-
4
Status
When the thresholds are exceeded, the service changes colour.
-
5
Incident
The outage is recorded with its start and end.
Reports
When someone picks a problem on a service page, we store a report: time, company and type of problem. Nothing else.
Reports are grouped into 15-minute intervals. This is the unit used in every chart.
One hour, four intervals
14:00
3
14:15
1
14:30
5
14:45
2
The usual level
Heavily used services receive some reports even when they are working normally. So what matters is not only the absolute number, but how far it moves away from the usual level.
For each company we calculate an hourly average over the last three weeks. This baseline varies by service and by time of day.
The same day, hour by hour
At three in the morning, a handful of reports can be unusual. At nine in the evening, the same number can be normal.
The three statuses
If reports in the last 15 minutes exceed two and a half times the baseline, the status changes to “Reports rising”. From five times the baseline it counts as an “Ongoing incident”.
An absolute minimum always applies as well, so that a few night-time reports cannot trigger an alert on small services.
Reports received are within the usual range for this service.
Volume is above the usual level, but not yet high enough to confirm an incident.
Report volume is well above average. An outage is probably under way.
Each bar represents 15 minutes. A minimum absolute number of reports is also required, so that two or three people cannot trigger an alert on little-used services.
The history
As soon as a service leaves the normal status, an incident begins. It only ends after a continuous quiet period.
We store the start, end, duration, report peak and breakdown by type of problem.
One incident instead of thirty alerts
Solid bars are above the threshold and light bars below it. The incident only ends after 30 minutes consecutive quiet minutes. This way a long outage is not split into many small alerts.
Limitations
ServiStat counts problem reports and does not run technical checks on servers. Small incidents may go unnoticed, and events that are widely discussed can generate extra reports.
The absence of reports does not guarantee that everything is working. Reports are not an official confirmation of an incident. For authoritative information, check the channels of the company concerned.
Last updated: 21 September 2026