Uptime SLA Explained: 99.9% vs 99.99% Downtime Compared

You are comparing two hosting plans. One promises 99.9% uptime, the other 99.99%. They look almost identical, right? That tiny extra "9" actually decides whether your website is offline for a long lunch break or a full working day every year. Let's unpack what those numbers really mean, in plain language, so you can choose with confidence.

What "uptime" actually means

Uptime is the percentage of time your website or server is up and reachable over a given period. If a month has 720 hours and your site was working for 719 of them, your uptime was about 99.86%.

The opposite of uptime is downtime: the minutes when visitors hit an error page, a spinning loader, or nothing at all. Think of uptime like a shop's opening hours. A shop that promises to be open "99.9% of the time" still locks its doors occasionally, you just hope it happens when nobody is knocking.

An SLA (Service Level Agreement) is the written promise your hosting provider makes about that number. It is part of your contract. It says, in effect: "We guarantee at least X% uptime, and if we miss it, here is what you get." The key word is guarantee, because an SLA is not just a marketing figure, it is a commitment with consequences (usually money back).

How much downtime each level allows

Here is where those nines stop being abstract. The table below shows the maximum downtime each uptime level permits. The yearly figures use a 365-day year; the monthly figures use a 30-day month. The math is simple: downtime allowed = (100% − uptime%) × total time.

Read that again. Moving from 99.9% to 99.99% shrinks your allowed yearly downtime from nearly 9 hours to under 53 minutes. That is the difference a single extra nine makes. And 99.999% ("five nines") is the gold standard, only about 5 minutes of downtime across an entire year, the level you would expect from a bank or an emergency service.

A quick rule of thumb to memorise: each extra nine cuts your allowed downtime by roughly 90%.

What counts as downtime (and what providers quietly exclude)

Here is the catch most people miss: the provider decides what "downtime" means in their SLA. Two providers can both promise 99.99% while measuring it completely differently.

Downtime usually means your service is fully unreachable or returning errors. But SLAs commonly exclude the following from the count:

1. Scheduled maintenance. Planned windows the provider announces in advance often don't count, even though your site is down for visitors.
2. Your own mistakes. A bad deploy, an expired SSL certificate, a plugin that crashes the site, or hitting your resource limits.
3. Problems outside their network. DNS issues, your domain registrar, your CDN, or the visitor's own internet connection.
4. Force majeure. Natural disasters, cable cuts, large-scale internet incidents.
5. Brief blips. Some SLAs only count an outage once it lasts longer than a minimum threshold (for example, 5 minutes continuously).

This matters because internet-wide disruptions are real and frequent. Cloudflare's 2025 year-in-review reported numerous national-scale internet outages and traffic disruptions across the year, many caused by power cuts, government-directed shutdowns, or cable faults, none of which a single host can fully control. Source: Cloudflare Radar 2025 Year in Review. Always read the definitions and exclusions section of an SLA, not just the headline percentage.

Monitoring vs the provider's SLA: why they disagree

You set up an uptime monitor (a tool that pings your site every minute) and it reports 99.7%. Your host insists they delivered 99.99%. Who is right? Often, both, because you are measuring different things.

1. Different vantage point. Your monitor checks from one location over the public internet. A network problem between your monitor and the server looks like downtime to you, but the server was technically "up."
2. Different definition. Your monitor counts every failed check. The SLA may ignore short blips, scheduled maintenance, or anything outside the provider's network (see exclusions above).
3. Different check frequency. A monitor pinging every 5 minutes can miss a 2-minute outage entirely, or count a single slow response as a full interval down.
4. The full chain. Visitors experience your whole stack: DNS, CDN, application, database. The SLA usually covers only the provider's slice of it.

The practical takeaway: run your own external monitoring so you have evidence. But expect your numbers to be stricter than the provider's, and know which definition the contract actually uses when you file a claim.

Credits vs real business impact

So what do you get when a provider misses its SLA? Almost always a service credit, a percentage of your monthly fee refunded as account credit, not cash. The worse the outage, the bigger the credit, but it is capped and it rarely reflects your real loss.

Imagine you pay 20 EUR a month for hosting. A bad outage might earn you a 25% credit, that is 5 EUR. But if that outage took down your online shop during a busy promotion, you may have lost far more in sales, abandoned carts, and trust. The credit covers the hosting bill, not your business.

A few realities worth knowing before you rely on an SLA:

1. You usually have to claim it. Credits are rarely automatic. You must report the outage, with evidence, inside a deadline (often 30 days).
2. Credits are capped. Most SLAs limit credit to a fraction of one month's fee, no matter how bad the outage.
3. No consequential damages. Lost revenue, reputation, or recovery costs are explicitly not covered.

So treat the SLA as a quality signal and a small safety net, not as insurance for your business. If downtime would genuinely hurt you, the right answer is higher uptime plus your own redundancy (backups, a CDN, a failover plan), not a bigger refund.

How to use this when choosing a host

1. Match the level to your risk. A hobby blog is fine with 99.9%. A shop, SaaS, or booking system should look for 99.99% or better.
2. Read the exclusions, not just the headline number.
3. Check how credits are claimed and what they are capped at.
4. Plan your own monitoring and a backup strategy regardless of the SLA.

FAQ

Is 99.9% uptime good enough for a website?

For a personal blog or a low-traffic brochure site, yes, 99.9% allows under 9 hours of downtime a year, which most visitors never notice. For a business that earns money online, aim for 99.99% or higher.

What is the real difference between 99.9% and 99.99%?

Allowed yearly downtime drops from about 8h 45m to about 52 minutes, roughly a 90% reduction. Each extra nine cuts the allowed downtime by about 90%.

Does scheduled maintenance count as downtime?

Usually not, if it falls inside an announced maintenance window. Your visitors still see a site that is down, but the SLA typically excludes it, so check the wording.

Why does my uptime monitor show worse numbers than my host's SLA?

Because you measure from a different location, with a different definition, and across the whole chain (DNS, CDN, app). The SLA covers only the provider's network and often ignores short blips and maintenance.

Will I get money back if my host misses its SLA?

You typically get a service credit (a percentage of your monthly fee) rather than cash, you usually have to claim it within a deadline, and it is capped. It will not cover lost sales.

Is 100% uptime ever guaranteed?

No honest provider promises true 100%. Even "five nines" (99.999%) allows about 5 minutes of downtime a year, because some failures are simply outside any single provider's control. Source: Cloudflare Radar 2025 Year in Review.

Ready to pick a host whose guarantee matches your needs? Compare providers by SLA in our service comparator.