99.9% uptime allows about 43 minutes of downtime per month. 99.99% allows about four. That is the entire arithmetic, and it is the least useful part of the promise.
The number only means something once you know three other things: what is being measured, who is doing the measuring, and what actually happens when the site goes down. Most uptime guarantees answer none of those.
The arithmetic nobody prints next to the badge
Percentages this close to 100 are hard to feel. Converted to minutes, they get honest fast. Assume a 30-day month.
- 99% — about 7 hours and 18 minutes down per month. Nearly a full working day.
- 99.9% — about 43 minutes per month, or roughly 8 hours and 45 minutes across a year.
- 99.99% — about 4 minutes per month.
- 99.999% — about 26 seconds per month.
Each extra nine costs roughly ten times more to deliver and removes 90% of the allowed downtime. That is why 99.9% is the number almost everyone advertises. It sounds like perfection and costs like ordinary hosting.
Note also what 99.9% permits: a single 43-minute outage in the middle of your busiest Tuesday is inside the guarantee. The month still passes. The badge still holds. You still lost the Tuesday.
A ping to the homepage is not a working website
Here is the part that matters more than the percentage. Most uptime monitoring checks whether a server returns a response to a request for the homepage. If something comes back, the site is counted as up.
A great deal can be broken while that check stays green:
- The checkout fails at the payment step, but the homepage loads.
- The contact form submits and quietly sends nothing.
- The homepage returns in 400 milliseconds while every category page takes 14 seconds.
- A plugin update breaks the booking calendar. Pages load. Nobody can book.
- An SSL certificate lapses. The server is up. Every visitor sees a browser warning.
- Search engines get a 500 error while cached visitors see a normal page.
In each case the uptime figure is untouched and the business is losing money. The percentage measures whether a machine answered the phone. It does not measure whether anyone got what they came for. If your site has ever felt broken while a status page insisted everything was fine, that gap is why — and it overlaps heavily with the slow-site problems we covered in why your website feels slow.
Who is holding the stopwatch
The second question is who reports the number. In most cases, the host reports on itself.
Self-reported uptime is measured by the provider's own tools, from inside or near their own network, against endpoints they choose, and averaged across their whole fleet rather than your one site. A platform can be at 99.98% across thousands of servers while the box your site sits on had a rough week. Fleet averages are real. They are just not about you.
External monitoring — a service outside the host's network, checking from several locations, on a short interval, watching more than the homepage — is a different measurement and often a less flattering one. It is also the only one that reflects what your customers experience.
Read the exclusions, because that is where the number lives
Uptime guarantees are contracts, and contracts have carve-outs. Scheduled maintenance is almost always excluded. So are outages traced to your own code, your plugins, your theme, a traffic spike, a denial-of-service attack, upstream network providers, or anything the host labels "beyond reasonable control."
Strip those out and a meaningful share of real-world downtime never counts against the guarantee. The 43 minutes are 43 minutes of a specific, narrow category of failure — hardware and platform faults the host agrees to own. Most of the outages that actually hurt small businesses sit outside that category. A bad plugin update at 9pm is not a data centre fault. It is still your site being down.
A credit is not a remedy
Suppose the guarantee is breached properly. The site was down for two hours, it was clearly the host's fault, and you noticed in time to file a claim within the window. What do you get?
Usually a service credit. Often a percentage of one month's hosting fee, applied to a future invoice, capped, and paid only if you ask.
Do the comparison. On a $30 monthly plan, a generous credit might be a few dollars. If those two hours cost you four orders, a booked consultation, or the one lead that was going to be a $12,000 project, the credit does not touch it. It refunds the rent on the shelf, not the stock that fell off it.
This is not fraud on the host's part. It is simply what an uptime guarantee is: a small, bounded refund of hosting fees. It was never insurance against lost business, and reading it as insurance is how people end up surprised.
Response and recovery beat the marketing figure
Downtime is not one number. It is a sequence, and each step is a place to lose or save time.
- Detection. How long before anyone knows? If the answer is "when a customer emails," everything downstream is already slow.
- Diagnosis. Is it the server, the database, a plugin, DNS, the certificate, or an upstream provider? Guessing wrong costs more time than the outage.
- Decision. Fix forward, or roll back to a known-good copy? Someone has to be willing to make that call.
- Recovery. Can a working version of the site be restored quickly, and has anyone confirmed the backup actually restores?
- Explanation. Afterwards, do you find out what happened and what changed so it does not repeat?
A host with a lower advertised percentage and a competent person who answers within the hour will cost you less downtime, in practice, than a host with an extra nine and a ticket queue. The nine is a claim about machines. The response is a claim about people.
This is the thinking behind how we run backups: twice-daily verified off-site copies with 30-day retention, kept off the server they came from, and checked so a restore is a routine action rather than an experiment. We wrote about the mechanics in backups, monitoring, and recovery. Monitoring exists so we find out before you do. When something does break, you email a person and get a same-day reply from an engineer, with emergency triage during business hours. If the fix falls outside the plan, it is $95 an hour and we quote it before starting.
What to ask instead of "what is your uptime?"
If you are comparing hosts, the percentage will be the first thing offered and the least informative. Better questions:
- What exactly do you monitor — just the homepage, or checkout, forms, and certificate expiry too?
- Who measures it, from where, and how often?
- What is excluded from the guarantee?
- If my site goes down at 10am on a weekday, who reads that email and how fast do they reply?
- How far back do backups go, are they stored off the server, and when did you last test a restore?
- If a plugin update breaks something, is fixing it included or billed?
Answers to those tell you what your bad day will look like. The percentage tells you what the marketing team settled on. If the answers are vague, that is the answer. Our take on what a real plan should cover is in maintenance and security, and if you are already in the middle of an outage, emergency website help is the faster door.
When the number is worth caring about
It is worth being fair here. Uptime figures are not useless. A host advertising 99% is telling you something real, and it is not good. A history of long, unexplained, repeated outages is a genuine signal, and so is a public incident log written like an adult wrote it.
What is not worth doing is paying more for an extra nine on a small-business brochure site or a shop doing steady but not enormous volume. The engineering that buys 99.99% — redundant everything, instant failover, staff watching screens around the clock — is real work and real money, and for most sites it solves four minutes a month while the actual risk sits in an unpatched plugin. Spend the money on maintenance, backups you have tested, and someone who answers. That is where the recoverable time is.
Nothing broke this month is a fine outcome. Knowing exactly what happens when something does is the better one.
Frequently asked questions
How much downtime does 99.9% uptime actually allow?
About 43 minutes per 30-day month, or roughly 8 hours and 45 minutes per year. 99.99% allows about 4 minutes a month, and 99% allows more than 7 hours a month. Each additional nine cuts the allowed downtime by 90% and costs substantially more to deliver.
Does an uptime guarantee compensate me for lost sales?
No. Almost every guarantee pays out as a service credit against future hosting fees, usually capped at a portion of one month and paid only if you file a claim in time. It refunds part of what you paid for hosting. It does not cover lost orders, leads, or bookings.
Why does Hosterr not publish an uptime percentage?
Because the figure would tell you less than it appears to. We would rather describe what we actually do: monitor the things that matter, keep twice-daily verified off-site backups with 30-day retention, and answer your email the same day with a real engineer. When something breaks, that is what shortens the outage.
