Website uptime: what the percentages really mean
Uptime is the share of time a website is reachable. Hosting companies love quoting it, usually with a lot of nines. The nines matter more than they look, because each one cuts the allowed downtime by a factor of ten.
What each percentage allows
| Uptime | Downtime per year | Per 30-day month |
|---|---|---|
| 99% | 3 days, 15 hours, 36 min | 7 hours, 12 min |
| 99.5% | 1 day, 19 hours, 48 min | 3 hours, 36 min |
| 99.9% | 8 hours, 45 min | 43 min |
| 99.95% | 4 hours, 23 min | 21 min |
| 99.99% | 52 min | 4 min, 19 sec |
| 99.999% | 5 min, 15 sec | 26 sec |
So "99.9% uptime" is a promise that the site can be down for most of a working day each year and still hit its target. For a personal blog that's fine. For an online shop, 43 minutes on the wrong Friday evening is real money.
How uptime gets measured
A monitoring service sends a request to the site on a schedule, commonly every 1 to 5 minutes, and records whether a healthy answer came back. Better services check from several countries at once and only count the site as down when more than one location agrees. That stops one bad network path from triggering a false alarm.
The interval matters. A monitor that checks every 5 minutes can miss a 3-minute outage completely, and it can't tell a 1-minute blip from a 4-minute one. Short checks are more accurate and usually cost more.
What counts as "down"
This is where uptime figures quietly differ. Some monitors only ask "did the server answer?" Others want a 200 code, or check that a specific word appears on the page. A site that loads an error message with a 200 code can look perfectly healthy to a lazy monitor while every visitor sees a broken page.
The fine print in uptime guarantees
When a host promises 99.9% in a service level agreement (SLA), read what's excluded. Scheduled maintenance usually doesn't count. Problems caused by your own code don't count. And the remedy for missing the target is typically a partial refund of that month's hosting fee, which won't cover much if the outage cost you sales.
Checking vs. monitoring
Our checker answers "is it down right now?" once, when you ask. Monitoring asks that question every few minutes, around the clock, and emails or texts you when the answer changes. If you run a website, you want monitoring. UptimeRobot, Better Stack and Pingdom are well-known options, and several have free plans for a handful of sites.
Whatever you choose, point it at a page that proves the site really works (like a product page), not only the home page. A home page served from a cache can stay up long after the rest of the site has fallen over.
Is a site down right now? Check it in a few seconds.
Check a website