Website Monitoring Dashboard for Publishers and Multi-Site Owners

What 99.9% Uptime Really Means in Minutes

Monitoring

October 29, 2026

Datacenter Server Racks (22370909788)

Image: Datacenter Server Racks (22370909788), Carl Lender from Sunrise, USA. CC BY 2.0, via Wikimedia Commons.

Here is an illustrative example: at 9:10 on a Tuesday morning, a shop owner we'll call Marcus gets an email from a customer who says the checkout page "has been down all morning." Marcus opens his hosting dashboard. It proudly shows 99.9% uptime for the month. Both things can be true at once, and the gap between them is worth understanding before you pick a host or set up monitoring.

The math behind the nines

Uptime is the share of time a site responds. The leftover is the downtime you're allowed. Using an average month of about 30.4 days, which is roughly 43,800 minutes:

Uptime Downtime per month Downtime per year
99% about 7 hours 18 minutes about 3.65 days
99.5% about 3 hours 39 minutes about 1.8 days
99.9% about 44 minutes about 8 hours 46 minutes
99.95% about 22 minutes about 4 hours 23 minutes
99.99% about 4.4 minutes about 53 minutes

So "three nines" leaves room for most of an hour of downtime each month. If that hour lands at 9 a.m. on your busiest weekday, it doesn't feel like 99.9% to the customer who hit it.

Think of it like a store that promises to be open 99.9% of the time. It could close for 44 minutes once a month, or for two minutes every day around lunch. Both keep the promise. Only one of them annoys regulars.

Your check interval decides what you see

A monitor only knows what it checks. If a tool checks your site every five minutes and the site goes down for three minutes between checks, the monitor may never see it. If the site is down during two checks in a row, the monitor will record roughly ten minutes, even if the real outage was seven.

That means two dashboards watching the same site can show different uptime, simply because they check at different intervals. Shorter intervals catch brief outages and give a more accurate number. Longer intervals smooth things over and look better on paper.

On DashCanopy, a tile's state comes from its most recent health check. We explain what online, degraded and offline mean on a site tile, including how slow responses are flagged before they become full outages.

Slow is a kind of down

Uptime figures usually count a site as up if it responds at all. A page that takes 15 seconds to load is technically up. To a shopper, it's broken. That's why response time matters alongside uptime. A month of 99.95% uptime with frequent slow spells can cost more sales than a month at 99.9% with one clean outage at 3 a.m.

What to do with this

You might be thinking you should just buy the host with the most nines. More nines usually cost more, and the number on a host's sales page is often a target in a service agreement, with credits if they miss it, not a guarantee your site never goes down. A credit on next month's bill doesn't bring back the sales you lost.

A few practical habits help more than chasing a number:

  1. Monitor from outside your host. A host's own status page can miss problems that only show up for visitors.
  2. Pick a check interval that matches the stakes. A store or a lead form deserves a shorter interval than a hobby blog.
  3. Alert a person, not just an inbox. An alert nobody sees for two hours turns a five-minute blip into a long outage.
  4. Write down what "down" means for you. The home page loading is not the same as checkout working.

When a tile does go red, our guide on what to do when a site tile goes offline walks through the first ten minutes.

Turning a percentage into minutes

The table above is the same arithmetic, written out. Take the minutes in the period and multiply by the slice that is not uptime. A non-leap year is 365 days, which is 525,600 minutes. One tenth of one percent of that year is 525.6 minutes, or 8 hours and 46 minutes, which is the yearly cell for 99.9 percent. A month of 30.4375 days is 43,830 minutes, and one tenth of one percent of that month is about 43.8 minutes, which the table rounds to about 44. Google's Site Reliability Engineering book treats a target expressed in nines as a product choice about how much unavailability you are willing to budget, in the chapter Embracing Risk. The chapter is about setting that budget, not about a monitoring product.

On DashCanopy the check interval is part of what the budget means in practice. Starter checks every 10 minutes, Pro every 5, and Agency every 1. An outage shorter than the gap between checks can end before the next request, so the tile never records it. A one-minute check can see a two-minute failure that a ten-minute check misses. The nines on a host status page and the nines on a tile are both percentages. They are not the same sample.

This is general guidance. Your host's service agreement defines what its uptime number covers.

Your next step: look up the uptime your host promises, convert it to minutes with the table above, and decide whether that much downtime at your busiest hour is acceptable.

Back to Blog