What does the 99.9% uptime figure actually cover

It covers availability of the hosting platform your sites answer from, measured by continuous monitoring rather than occasional sampling. It is a statement about the infrastructure between the internet and your account. It is not a statement about your application, your DNS, your registrar or anything you installed.

Inside the figure

• The server your account lives on, reachable and serving.
• The network path to it, including the address your site answers on.
• The platform services your site depends on, such as the web server and database engine.

Outside it

Being clear about the boundary is more useful than a larger number would be:

• Your application. A plugin update that takes the site down is downtime for your visitors and not a platform outage. The distinction feels academic until you are the one debugging at midnight.
• DNS you control. A record pointed at the wrong address makes the site unreachable while the platform is entirely healthy.
• Your registrar. An expired domain or a delegation problem sits with them.
• Scheduled maintenance, where announced in advance.

What the number means in time

Three nines is roughly forty minutes a month. That is small, but it is not zero, and the shape matters more than the total. Forty minutes spread as a few seconds here and there is almost invisible. The same forty minutes in one block during a crawl is a different experience entirely. Continuous monitoring exists to catch the second kind quickly rather than to make the arithmetic look better.

Why continuous monitoring is the operative word

A provider that checks hourly can miss a ten-minute outage completely and report a perfect month in good faith. Continuous checking catches short interruptions, which are the ones that quietly accumulate across an estate. It also means the response starts before you notice, which is the practical benefit rather than the reporting one.

How to hold anyone to a figure like this

1. Run your own monitoring from outside. An independent check is the only version you can act on, and it also tells you about failures in your own application that a platform figure never will.
2. Alert on your domain, not the server, so DNS and certificate problems surface too.
3. Keep the history. A pattern over months is worth more than any single incident.

Support runs continuously, which is what makes a short interruption actionable at an awkward hour rather than something you report the next morning. Backups and restores are covered in Uptime and Backups, and the guarantee terms are on why choose us. The wider question of separating a host problem from a site problem is in Diagnostics.