How to Check a Web Host's Real Uptime Before You Buy
Quick answer: The uptime figure on a hosting sales page is self-reported and works as a refund rule, not as proof of service. Before paying, do three checks: read the host's status page history for the last six to twelve months, put a free uptime monitor on sites they already host, and buy one cheap month to test a real install and a late-night support ticket. Convert the percentage into minutes, because 99% means about seven hours offline every month.↗ Share on X
The "99.9% uptime" printed on a hosting sales page is not a measurement. It is a refund rule. Nobody is checking it for you, and in most contracts it only means that if the host goes below that number *and you notice, and you open a ticket, and you ask*, you get a small credit on your next invoice. So do not buy based on that number. Instead, do three checks that take about thirty minutes total: read the host's own status page history, put a free monitor on a site they already run, and run a short paid test with your own domain before you move anything important.
Here is exactly how to do each one, and how to read what you find.
What does 99.9% uptime actually mean in minutes?
Shared vs VPS Hosting: 9 Signs You Picked the Wrong Plan →
Is Your Web Host Killing Your Sales? A 30-Minute Test →
Switch Web Hosts Without Downtime: 9-Step Business Plan →Percentages hide the damage. Convert them into time and the differences get very loud.
| Uptime figure | Downtime per month | Downtime per year |
|---|---|---|
| 99% | about 7 hours 12 minutes | about 3 days 15 hours |
| 99.5% | about 3 hours 36 minutes | about 1 day 20 hours |
| 99.9% | about 43 minutes | about 8 hours 45 minutes |
| 99.95% | about 22 minutes | about 4 hours 23 minutes |
| 99.99% | about 4 minutes | about 53 minutes |
Two things jump out of that table.
The gap between 99% and 99.9% is enormous: seven hours a month against forty-three minutes. A host advertising "99%" is telling you it is fine to be offline for most of a working day, every month.
And the figure alone does not tell you the shape of the outage. Forty-three minutes spread over eight tiny blips is an annoyance. Forty-three minutes in one block, on a Monday morning, during a promotion, is a real loss. When you look at history, you care about the shape as much as the total.
Smart software picks in your inbox
Why the number on the sales page means so little
Three reasons, and it is worth knowing all of them before you talk to a salesperson.
It is self-reported. The host measures itself, from its own network, usually by checking whether the server responds, not whether your site loads correctly. A server that answers while your database is down counts as "up."
It usually excludes the outages that matter. Read the terms and you will find "scheduled maintenance" excluded. Scheduled by whom? By them, sometimes with a few hours of notice. Those hours do not count against the number.
It is a payout term, not a promise of service. This is the part most people miss. Most contracts say that if uptime falls below the stated figure, you may request a credit worth a fraction of one month's fee. On a cheap plan, that is a couple of dollars. It will not cover what you lost.
So treat the advertised figure as the floor of what they are willing to be held to, and go find the evidence yourself.
Check 1: read their status page history
Cheap Web Hosting: 11 Mistakes That Cost You More Later →
Web Hosting Cost for a New Business Site: Do the Real Math →
Shared vs VPS Hosting: Which One Does Your Site Need? →Almost every host has a status page. Look for a link in the footer, or search for the host's name followed by "status page". Many are hosted on public incident platforms.
Do not look at today. Look at the last six to twelve months, and answer four questions:
1. How many incidents are there? A page with zero incidents in a year is not reassuring, it is suspicious. Everyone has incidents. A page with none is a page nobody updates.
2. How long did they run? Note the start and end timestamps. Incidents that regularly run over four hours point at thin staffing outside business hours.
3. Do the same components keep breaking? Three separate incidents on the same database cluster is a structural problem, not bad luck.
4. Are the updates written by a human? Real postmortems say what broke and what they changed afterwards. Repeated copy-paste of "we are investigating an issue" and then "resolved" tells you they will not explain anything to you either.
If you cannot find a status page at all, that is a finding in itself. It means when your site goes down, you will have no way to tell whether it is you or them, and your only option is a support ticket.
Check 2: monitor a site they already host, for free
You do not need to be a customer to measure a host. Do this:
1. Find a site already hosted with them. Look at their own homepage, or the customer showcase they publish, or search for the host's name and "customers".
2. Sign up for a free uptime monitoring service. The free tiers usually allow a handful of checks every five minutes, which is plenty.
3. Add two or three of those sites, plus the host's own control panel login page. The control panel matters: if it is down while your site is up, you cannot fix anything.
4. Let it run for two weeks and read the report.
Two weeks will not give you a statistically solid annual figure, and it does not need to. What you are looking for is obvious trouble: daily blips, slow response creeping up in the evening, or an outage that the status page never mentioned. Any of those three tells you enough.
Check 3: run a short paid test before you move anything
Never migrate a working site as your first act with a new host. Buy the cheapest month, and test with something disposable.
1. Buy one month on the plan you are considering, not the promotional three-year deal. The long deals are cheap because they remove your ability to leave.
2. Point a spare domain, or a subdomain you control, at the new host.
3. Install the same software your real site runs, and load it with a realistic amount of content, not an empty homepage. Empty sites are fast everywhere.
4. Measure the time to first byte from two or three different countries using any free speed testing tool. Under 400 milliseconds is good on shared hosting. Consistently over 800 milliseconds is a warning.
5. Open a support ticket on purpose. Ask a real question, at an awkward hour, like 11 p.m. on a Saturday. Time the first reply and read it. Did a human answer, or a script? This single test predicts more of your future pain than any uptime figure.
6. Ask support directly, in writing: "What is your backup retention, and how do I restore a single file from three days ago by myself?" If restoring requires a ticket and a fee, factor that in now.
Read the refund clause before you believe the promise
Find the terms of service and search for "uptime" or "SLA", which stands for service level agreement, the contract section that defines what they owe you if service falls short. Then check four specific things:
- Who measures it? If it is them, with no third-party reference, the number is theirs to define.
- Is scheduled maintenance excluded? Almost always yes. Ask how much notice they give.
- Do you have to request the credit? Usually yes, and usually within a short window like thirty days. Credits are almost never automatic.
- What is the maximum payout? Often capped at one month of fees. That is the real value of the promise.
None of this makes a host bad. It makes the advertised number close to meaningless as a buying signal, which is exactly why the three checks above matter.
Red flags to walk away from
- No public status page anywhere
- Support that only exists through a ticket form, with no stated response time
- A discount that only applies to a multi-year prepayment, with renewal at three or four times the price
- No self-service backups, or backups sold as a separate paid add-on
- Terms that mention unlimited storage and bandwidth but include a vague "fair use" clause with no numbers
- No documented migration path out, which is a good sign they expect leaving to be painful
Your next step
Before you pay anyone, do this in the next twenty minutes: open the host's status page and read the last six months of incidents, then add two of their customer sites to a free uptime monitor.
Set a reminder for two weeks from now to read that monitoring report. Then buy one month, not three years, and send a support ticket late on a weekend before you move a single file.
FAQ
How much downtime does 99.9% uptime actually allow?
About 43 minutes a month, or roughly 8 hours 45 minutes a year. By comparison, 99% allows about 7 hours 12 minutes every month, and 99.99% allows about 4 minutes a month. The shape matters too: one long block on a Monday morning hurts far more than several short blips.
Can I measure a host's uptime before becoming a customer?
Yes. Find sites they already host, including their own homepage and control panel login page, and add them to a free uptime monitoring service. Let it run for two weeks. You are looking for daily blips, response times creeping up, or outages the status page never mentioned.
What should I look for in the uptime clause of the contract?
Check who measures it, whether scheduled maintenance is excluded, whether you have to request the credit yourself within a time limit, and what the maximum payout is. Most contracts cap it at a fraction of one month's fee, which is why the advertised number is a weak buying signal.