Uptime tells you how much of a period a website was working. If a site was available for 29 days and 23 hours of a 30-day month and unavailable for the remaining hour, its uptime for that month is about 99.86%.
The word is often used as a synonym for availability. In everyday use they mean the same thing: was the site there when someone wanted it? Engineers sometimes keep a distinction, using availability for the outcome a user experiences and uptime for the time a system is running. For a website owner the practical question is the same: for how much of the time could visitors use the site?
How uptime is calculated
The formula is simple:
uptime % = (total time − downtime) ÷ total time × 100
Take a 30-day month, which has 43,200 minutes. Suppose the site had two outages, one of 12 minutes and one of 31 minutes. Downtime is 43 minutes.
(43,200 − 43) ÷ 43,200 × 100 = 99.90%
Two things decide whether such a number is honest. The first is what counts as "down" (see downtime). The second is how the downtime was measured, which is covered below.
What the percentages mean in minutes
Uptime percentages look close together and are not. Each extra "nine" cuts the allowed downtime by a factor of ten.
| Uptime | Downtime per 30-day month | Downtime per year |
|---|---|---|
| 99% | 7 hours 12 minutes | 3.65 days |
| 99.9% | 43 minutes 12 seconds | 8 hours 45 minutes 36 seconds |
| 99.99% | 4 minutes 19 seconds | 52 minutes 34 seconds |
A site at 99% can be unavailable for most of a working day each month and still report a number that looks high. At 99.99%, a single slow restart can use up the entire month's allowance.
Uptime and downtime
Uptime and downtime are two sides of the same measurement. Whatever is not uptime is downtime, so a 99.9% month has 0.1% downtime. Reporting uptime puts the focus on the healthy share, which makes small problems look small. Reporting downtime in minutes and counting incidents often tells you more about what visitors actually experienced.
Where uptime figures come from
A monitoring service does not watch a site continuously. It sends a request at a fixed interval, such as every five minutes, and records the result. Uptime is then estimated from those samples.
This has consequences:
- Short outages can fall between checks. With a five-minute interval, an outage of two minutes may start and end without a single failed check. A shorter interval narrows the gap but does not remove it.
- Durations are approximate. An outage seen by two consecutive failed checks is known to have lasted somewhere around the interval, not to the second.
- Different tools give different numbers. Two services that check from different places, at different intervals, and with different rules for what counts as a failure will report different uptime for the same site.
For that reason, an uptime figure from a monitoring tool describes what the tool observed under its own rules. It is not a measurement of the site from every visitor's point of view.
Why uptime matters
When a site is down, visitors cannot buy, sign in, read, or contact you. The cost depends on the site. For a shop, it is lost orders. For a software product, it is customers who cannot use what they pay for. For a brochure site, it is a missed first impression.
Uptime also matters because it is easy to promise and measure. Hosting providers, agencies, and software vendors state uptime targets in contracts, which is where terms such as SLA come from.
What uptime does not tell you
- A site can be "up" and still unusable. A server may answer with a success code while the checkout, login, or search is broken. Uptime based on web requests sees the answer, not the experience.
- Slow is not down. A page that takes twelve seconds to respond is available by most definitions. Response time tracks that separately.
- Averages hide patterns. A month at 99.9% may be one 43-minute outage or thirty short ones. Customers experience those very differently.
- A high number does not guarantee anything. Past uptime says nothing certain about the next month.
How to use the number
Pick a period and keep it constant, usually a calendar month. Track downtime in minutes alongside the percentage. Look at individual incidents rather than only the total. And treat a target such as 99.9% as a measurable goal, not as a promise that nothing will fail.