In short
99.9% uptime allows about 8 hours 45 minutes of downtime per year, or roughly 43 minutes a month. 99.99% allows about 52 minutes a year; 99.5% allows almost 44 hours. An uptime guarantee in a hosting contract is a service-level agreement: if the provider misses it, you typically receive a credit worth a fraction of the monthly fee, not compensation for lost business. For most sites the more useful questions are how quickly outages are detected and fixed, and whether the downtime lands during your busy hours.
The arithmetic
Uptime percentages translate into allowed downtime like this:
| Uptime | Downtime per year | Per month | Per week |
|---|---|---|---|
| 99% | 3 days 15 hours | 7 hours 18 min | 1 hour 41 min |
| 99.5% | 1 day 20 hours | 3 hours 39 min | 50 min |
| 99.9% | 8 hours 46 min | 43 min | 10 min |
| 99.95% | 4 hours 23 min | 22 min | 5 min |
| 99.99% | 52 min | 4 min | 1 min |
| 99.999% | 5 min | 26 sec | 6 sec |
Each additional nine cuts the allowance by ten. Each one also costs roughly ten times as much to engineer, which is why consumer hosting promises three nines, serious platforms promise four, and five is the territory of telephone networks and payment rails.
What the guarantee actually promises
An uptime figure in a hosting contract is a service-level agreement, and it is worth reading the definitions rather than the headline.
What counts as down. Usually only a complete failure to respond, as measured by the provider's own monitoring. A site that responds in thirty seconds, returns errors for half of requests, or is unreachable from one continent is often "up" by the contract's definition.
What is excluded. Scheduled maintenance, almost always. Problems caused by your own code or configuration. Attacks. Upstream network failures. "Force majeure." The list of exclusions is frequently longer than the guarantee.
What you get if it is missed. A service credit: typically 5–25% of that month's fee, claimed by you within a deadline, applied to a future invoice. If the site earns $2,000 a day and is down for a day, the credit on a $30 plan is a few dollars. No hosting SLA compensates for lost revenue, and none will.
So the number is a statement of the provider's confidence and engineering, not an insurance policy. Treat it as the former.
Why the percentage is the wrong thing to optimise
Two hosts each deliver 99.9%. The first has one four-hour outage a year at three in the morning on a Sunday. The second has thirty ten-minute outages spread across weekday afternoons. Same number, very different businesses.
The measures that describe the experience better:
Time to detect. How long the site is down before anyone knows. Without external monitoring, the answer is often "until a customer emails", which can be hours. A free uptime checker pinging the site every minute reduces this to a minute.
Time to recover. How long from detection to fixed. This is where hosting quality actually shows: a managed platform that fails over automatically recovers in seconds; a cheap shared host with a support queue recovers when someone reads the ticket.
When it happens. Downtime during your peak hours costs many times what the same downtime costs at night. If a provider's maintenance windows are your busy periods, the percentage is irrelevant.
Partial degradation. Slow, erroring, or half-working is not counted as downtime by anyone's SLA and is experienced as downtime by every visitor. Monitoring a production app properly means watching error rates and response times, not just whether the server answers.
What different sites actually need
A static marketing site on a CDN gets effectively 99.99%+ without anyone trying, because there is no server to fail; files are replicated across many locations and the failure of any one is invisible. This is one of the quieter arguments for static sites.
A small business site on shared hosting typically experiences 99.5–99.9% in practice, with the occasional multi-hour outage from a neighbour's problem or a maintenance overrun. For most, that is acceptable, provided the site is monitored and someone can act.
An application people pay for needs 99.9% as a floor and, more importantly, fast recovery: health checks, automatic restarts, a database with a replica, and a deploy process that can roll back in minutes. Reaching this is mostly a matter of not doing the things that cause outages: deploying at peak, running without backups, having no alerting.
Anything where minutes cost real money needs an engineered answer with redundancy at every layer, a tested failover, and a team on call. That is a budget line, not a hosting plan.
Your own uptime, not theirs
The host's uptime and the site's uptime are different numbers. The host can be perfect while the site is down because a deploy broke it, a certificate expired, a domain lapsed, a plugin update failed, or the database filled its disk. For most small sites these self-inflicted outages are at least as common as host outages, and none of them appears in any SLA.
The practical programme for a site that matters: external monitoring with alerts that reach a phone; certificate and domain expiry on a calendar; backups tested by actually restoring one; a deploy process that can be reversed; and someone who knows they are responsible. That produces better real-world availability than moving to a host with an extra nine.
If you would like to know what your site's actual availability has been, rather than what the contract says, book a call and we will set up the monitoring that tells you.
Common questions
How many hours of downtime is 99.9% uptime?
About 8 hours and 46 minutes per year, or roughly 43 minutes per month. 99.99% allows about 52 minutes a year; 99.5% allows almost 44 hours. Each additional nine reduces the allowance tenfold and costs roughly ten times as much to engineer.
Is 99.9% uptime good?
For a marketing site or a small business site, yes; it is the standard offer from reputable hosts and the practical experience is usually better. For an application customers pay for, it is a floor rather than a target, and the more important measures are how quickly outages are detected and fixed and whether they fall during busy hours.
What happens if a host breaks its uptime guarantee?
You can usually claim a service credit, typically 5–25% of that month's fee, applied to a future invoice, provided you claim within the contract's deadline and the outage was not excluded as maintenance, your own fault, or an attack. No hosting SLA compensates for lost sales or reputation.
What is the difference between uptime and availability?
Uptime measures whether the server responds at all. Availability, properly measured, is whether the service is usable: responding quickly, without errors, for all users. A site can be 'up' while returning errors or taking thirty seconds per page. Monitoring response times and error rates gives a truer picture than a ping.
How do I monitor my website's uptime?
Use an external uptime service that requests a page every minute from several locations and alerts you by email, SMS or app when it fails or slows. Several are free for a handful of checks. Monitor a page that exercises the real application, not just the homepage, and put certificate and domain expiry dates on the same alerting.
