The uptime we measure,
in the open.

How this page is produced.

The probe

Each monitor is a plain HTTPS check against the endpoint an application would call. That is an API root where one is public, the primary hostname otherwise. We never probe a vendor's own status page, because that measures their status provider.

The threshold

Checks run every 60s. Three consecutive failures open an incident and flip the component. Two consecutive successes close it. A single blip never counts as downtime.

The caveat

One vantage point on the public internet. It sees what a user in that region sees. A regional outage elsewhere may not show up here, and these numbers have nothing to do with any vendor SLA.