Blog / How the internet works /

Uptime vs availability: what users actually feel

Your monitoring tool shows a cheerful green line and 100% uptime. Meanwhile a customer emails to say they’ve been unable to check out for an hour. Both can be true. Uptime tells you whether the server answered. Availability asks whether people could actually do what they came to do. The gap between the two is where unhappy users live.

The definitions, simply

Uptime is the share of time a service responds at all. A check pings your site, gets a reply, and counts it as “up”.

Availability is the share of time the service is usable for the thing that matters. It can be measured at different levels: the home page loads, search works, checkout completes, the login succeeds.

Think of a restaurant. The lights are on and the door is open (uptime). But the kitchen is closed and the only thing they can serve is water (not really available). A sign at the front would say “open”. A customer would say “not really”.

How a site is “up” but unusable

  • The home page loads, checkout fails. The server answers, but the payment step errors out.
  • Pages load painfully slowly. Technically responding, but after ten seconds most people give up.
  • A broken script blocks the page. The HTML arrives, but a failing script leaves buttons dead or a popup covering everything.
  • Search or the basket is broken. Core features fail while static pages work.
  • Wrong content is served. A cache or deployment error shows an old or empty page with a “200 OK” status.
  • Only some regions or networks can reach it. Your check from one location says fine, while readers elsewhere see errors. A cut cable or routing problem can do this, as I described in how undersea cable cuts slow down countries.
  • An error page that returns a normal status. Monitors read the success code and say all’s well.
  • Third-party failures. The payment provider, the CDN or a font service is down, taking part of your site with it.

None of these show up as “down” in a simple check.

What percentages look like

Uptime is often quoted as a percentage, and the figures sound similar while meaning very different things over a month. Roughly:

Uptime Allowed downtime per 30-day month
99% about 7 hours 12 minutes
99.9% about 43 minutes
99.99% about 4 minutes 19 seconds

(These come from simple arithmetic: 30 days is 43,200 minutes.)

Two things worth noticing. First, “99%” sounds excellent but allows most of a working day of downtime each month. Second, the figure says nothing about when. Forty-three minutes in the middle of the night is a shrug. Forty-three minutes during a sale is a disaster. For the money side of this, see website downtime: what it really costs a small business.

Also read the small print on a hosting promise. A “99.9% uptime guarantee” usually refers to the server or network being reachable, often compensates only with account credit, and excludes planned maintenance. It’s not a promise about your whole site working.

How to measure what users feel

Move from “does it answer?” to “does it work?”

  1. Check content, not just status. A good monitor looks for a particular word on the page or a specific response, so an empty page or error message trips it.
  2. Monitor key journeys. Some tools can run a scripted path: load a product page, add to basket, reach checkout. If you can’t, test it by hand on a schedule.
  3. Check from several locations. One vantage point hides regional problems.
  4. Track speed as well as presence. A response time that doubles is a warning, even if nothing is “down”.
  5. Watch real users. Analytics and error reports show drops in conversions, spikes in errors or sudden falls in traffic that no ping would catch.
  6. Listen to people. Support emails, social mentions and a plain “report a problem” link are a monitoring system too.
  7. Monitor your dependencies. Know where your payment, email and CDN providers publish their status, and watch them.

My own basic monitoring setup is in the uptime monitoring setup I use for client sites. The ideas here are the next layer on top of it.

Set targets people can feel

Instead of a single percentage, try targets stated like a user would state them:

  • “Customers can complete checkout 99.9% of the time.”
  • “Pages load in a few seconds for most visitors.”
  • “Contact forms always deliver, and we are alerted if one fails.”

It forces you to decide what really matters, then measure it. Some engineering teams call this a service level objective. For a small site, the plain-English version is enough.

What to do when “up” lies to you

  • Add a keyword check to your monitor.
  • Test checkout and forms after every update.
  • Set an alert on sudden falls in orders or form submissions.
  • Keep a status note or social channel ready for when things go wrong.
  • After any incident, ask “how would we have known sooner?” and add that check.

Key takeaways

  • Uptime means the server answered; availability means people could do what they came to do.
  • A site can be “up” while checkout, search, or an entire region is broken.
  • Percentages hide timing: a short outage at a busy moment costs more than a long one at 3 a.m.
  • Monitor content, journeys, speed and multiple locations, not just status codes.
  • Define targets in terms of what users experience.

What is the one thing a visitor must be able to do on your site? Is anything checking it?