Uptime, downtime and what an SLA guarantee actually promises

“99.9% uptime” sounds close enough to perfect that most people never do the arithmetic on it. Over a full year, 99.9% still permits roughly nine hours of downtime. 99.5% — also commonly advertised — permits closer to a full working day and a half. Neither figure describes when that downtime happens. Both only say how much of it is allowed before the guarantee is technically breached.

What an uptime guarantee’s SLA percentage actually is

An uptime guarantee is a promise about the proportion of time, measured over an agreed period, that the host’s infrastructure will be available. It says nothing about how that unavailable time is distributed. A hosting provider meeting a 99.9% guarantee could, in principle, deliver it as one continuous nine-hour outage during a business’s busiest week of the year, and still be technically compliant. This promise is usually the headline term of a wider service level agreement, or SLA. That agreement also covers the service credit and the process for claiming it — both covered below.

Why hosts talk about “nines,” and what “five nines” means, in minutes a year

Hosting providers often shorthand an uptime guarantee by how many nines the percentage contains. 99% is “two nines” and permits roughly 3.65 days of downtime a year. 99.9% is “three nines” and permits roughly nine hours. 99.99% (“four nines”) permits roughly 52 minutes. 99.999% (“five nines”) permits only around five minutes a year. Five nines is the figure associated with the most demanding infrastructure, telecommunications networks for example. It is rarely what an ordinary small-business hosting SLA actually offers or needs. Knowing which “nines” tier is being quoted turns a marketing percentage into an actual number of minutes. That is the only version of the figure worth comparing.

Why the uptime percentage itself rarely differentiates one host’s downtime from another’s

Nearly every reputable host advertises a similar figure, generally somewhere between 99.9% and 99.99%. That makes the number itself a poor basis for choosing between them. Everyone is claiming roughly the same thing. What actually varies, and what is worth comparing directly, is what happens around an outage, rather than the round number attached to preventing one.

What actually matters in the service level agreement: the service credit, and whether it is worth claiming

Most hosting agreements attach a service credit to a missed guarantee, commonly a partial refund of that billing period. Sometimes the customer has to actively notice the breach and formally claim it within a limited window. A credit that has to be manually claimed, against a threshold most businesses never bother tracking closely enough to notice has been crossed, is a guarantee that mostly protects the provider’s reputation. It does not compensate the customer in any meaningful way.

What matters more: communication during an outage

A host that posts to a genuinely independent, regularly updated status page the moment an issue begins, states plainly what is affected, and updates as the situation changes, is offering something considerably more valuable than a compliant percentage on paper. A host whose customers first discover an outage by noticing their own site is down is a materially worse experience. There is no acknowledgement anywhere for an extended period, regardless of what the annual uptime figure eventually works out to.

What to actually ask your hosting provider before signing up

Ask what the service credit process actually requires: automatic, or does the customer have to notice and claim it? Ask where the status page lives, and whether it is genuinely updated during an incident, rather than only after the fact. And, separately, ask whether outages are more commonly caused by the host’s own infrastructure or by scheduled maintenance windows. Planned maintenance handled with adequate notice is a different and far less concerning pattern than unplanned failures.

What uptime guarantees cannot protect against

An uptime guarantee covers the host’s own infrastructure. It says nothing about a site being unreachable because of an expired domain, a misconfigured DNS record, or a compromised installation taken down deliberately as a containment measure. All of these are outages by any practical definition. None of them are covered by any hosting uptime guarantee, because none of them are the host’s fault.

Independent monitoring, briefly

A third-party uptime monitoring service checks the site independently of the host’s own reporting. It gives a business its own record of actual availability, rather than relying entirely on the host to self-report accurately. It is a genuinely useful, low-cost addition, regardless of how strong a host’s own guarantee reads on paper.

What a genuinely bad pattern looks like over time

A single brief outage, promptly communicated and resolved, is unremarkable and not a reason to change host. A recurring pattern of outages, or a host that stops communicating clearly once incidents become frequent, is the actual signal worth acting on. The difference between the two is pattern, not any single event.

Why this page exists separately from a general host comparison

Uptime is the single most repeated marketing figure in the entire hosting market. Precisely because it is repeated so uniformly, it deserves the specific, sceptical treatment given here, rather than being accepted as a meaningful differentiator on its own.

Scheduled maintenance, briefly

Planned maintenance windows, communicated well in advance and scheduled during genuinely low-traffic periods, are a normal and reasonable part of running infrastructure. They are worth distinguishing clearly from unplanned outages when assessing a host’s actual reliability. Treating the two as equivalent unfairly penalises a host that is simply being transparent about routine work.

Reading a status page critically

Some hosts run their status page on the same infrastructure the page is meant to report on. That means a total outage can take the status page down with it. A genuinely independent status page, hosted elsewhere, is a stronger signal of a host taking transparency seriously than one that could disappear at exactly the moment it matters most.

Keeping a simple incident log

A short, ongoing record of any outage a business notices, however brief, builds an independent history. That is genuinely useful the day a pattern needs to be demonstrated to a host, rather than relying on memory alone.

Where to go from here on the availability a hosting promise actually delivers

Whether the underlying infrastructure genuinely suits the site’s needs in the first place is set out in shared hosting versus VPS versus managed hosting. Ongoing monitoring that catches an outage regardless of whose fault it is, rather than relying purely on a host’s own reporting, is covered in what website maintenance costs.