6 min read · Updated November 20, 2026

How Outage Detection Works: Probes, Status Pages and User Reports

What really happens when a site checks if a service is down – and why combining live tests, official status and user reports gives the most reliable answer.

When millions of people suddenly can’t reach an app, outage trackers light up within minutes. But how do they actually know a service is down? There are three main methods, each with strengths and blind spots. DownFromSpace combines all three.

1. Active monitoring with probes

A probe is a small computer somewhere in the world that sends a request to a service and measures the answer. If a probe in Germany asks for youtube.com and gets a normal response in 80 milliseconds, YouTube is reachable from that network at that moment.

One probe proves little – it might have its own network problem. That is why we use probes in 50 countries at once, from the open Globalping network. When dozens of probes on different continents fail together, the problem is almost certainly the service itself. When only one region fails, the cause is often a regional data centre, an internet exchange problem or a government block.

Each result tells us three things:

  • Did the server answer? No answer or a timeout means it is unreachable from there.
  • What status code did it return? 5xx codes mean the server itself has a problem. Codes like 403 or 405 often just mean the service doesn’t like automated requests – it is still online. Learn more in our HTTP status code guide.
  • How fast was it? A very slow answer is often the first sign of an overloaded service, minutes before it fails completely.

2. Official status pages

Most large tech companies run a public status page that lists incidents and maintenance. Many use a standard format that machines can read, so we can show the current official state directly on each service page.

Status pages are valuable because they come from the company that actually sees its own servers. Their weakness is timing and honesty: they are updated by people, often after users have already noticed, and a huge outage is sometimes described as “some users may experience issues”.

3. User reports

Crowdsourced reports are the third signal. A probe can see that a website loads, but not that logging in fails, that a payment is rejected or that a game’s matchmaking never finds a match. Users notice those problems immediately.

Reports need careful handling. A single person with a broken router can report a “total outage”. That is why reports matter most in volume – a sudden spike compared with the normal level – and why we limit each visitor to one report per service every 30 minutes.

Putting it together: the Down score

On every service page we combine the three signals into one Down score from 0 to 100. It takes the highest of three values: the share of failing countries in the latest global test, the severity of the official status and the number of user reports in the last hour. A score near 0 means everything looks normal; above 60 means a serious problem.

The limits of any outage tracker

  • Probes mostly sit on data-centre and major networks, not on every home internet provider.
  • Some services are blocked in certain countries – China, Russia and Iran block several Western services – so red there can be normal.
  • A website can be up while its app, login system or game servers are down.

That is why we always recommend combining the global test with your own “Is it just me?” check. If the world is green and your own test fails, read our guide Is it down or just me?

Advertisement

Check a service now

More guides