Website Uptime Monitoring: What We Check Before Your Visitors Do
Your website can be offline for a week and you will not get one message about it. Hosting dashboards show green, your inbox stays quiet, and the first person to tell you is a customer you wanted to impress. The fix is a small one: website uptime monitoring with free tools, about twenty minutes of setup, and one check a plain status monitor misses.
How long a site can stay down unnoticed
We found the gap on our own infrastructure. On 30 September 2026 we pointed a status sweep at seven URLs we operate, including track.dev2production.com, the pixel tracker we built for counting email opens. The domain answers in about 50 milliseconds over a valid TLS connection, which is exactly what most status pages call healthy. What it hands back on the homepage is a 404. Anyone typing that address into a browser gets an error page instead of a site. The deployment never flagged it because health checks run against the routes the service exists for, and the homepage was never one of them.
An uptime monitor closes the gap between “the service works” and “a visitor can use it”. The good news is the price. UptimeRobot’s free plan watches 50 URLs at five-minute intervals and alerts by email, SMS or phone app push, no card required, and its paid tiers check every 60 seconds. Uptime Kuma is the self-hosted alternative: MIT-licensed, 92,000 stars on GitHub, running as one container on a server you already pay for.
What Google does while your site is down
An outage costs more than the visits that bounce. Google’s crawler documentation, last updated in February 2026, states that URLs returning 404 and other 4xx codes are removed from the index, and that URLs persistently returning 5xx errors are dropped too, after crawling has been slowed down. A 429 counts as a server error, not a client error.
So an outage that lasts days does damage that outlasts it. Pages you spent months getting ranked quietly leave the index, and recovery means waiting for the crawler to return and trust the site again. Turning an outage from days into minutes is, for a business that lives on search, the difference between a hiccup and a hole in the pipeline. That is what a five-minute monitor buys.
What we check, twice, before a build is called done
Before any site of ours goes live we run the sweep an uptime service will run later: seven URLs, two passes half a minute apart, and for each one the exact status code, transfer size and time to first byte.
| URL | Status | Bytes | First byte, pass 1 | First byte, pass 2 |
|---|---|---|---|---|
| dev2production.com | 200 | 185,471 | 0.61 s | 0.20 s |
| agents.d2p.site | 200 | 30,116 | 0.52 s | 0.13 s |
| warasat.com.pk | 200 | 64,526 | 0.35 s | 0.08 s |
| mtufaildiagnosticcenter.com | 200 | 10,317 | 0.20 s | 0.07 s |
| truckaxic.com | 200 | 118,570 | 1.91 s | 0.38 s |
| media.d2p.website | 200 | 240 | 0.05 s | 0.04 s |
| track.dev2production.com | 404 | 207 | 0.05 s | 0.04 s |
The customer who finds the outage first does not send a report; they send their money somewhere else. The metric that matters is time to detect, and the whole job of a monitor is to hold it under five minutes.
The four monitors a small site actually needs
UptimeRobot lets you define 50 monitors on the free plan, which is far more than a small business needs, so the work is choosing the right few. We set up four kinds on a client site.
First, the homepage. It is the URL customers type and the one Google recrawls most. Second, the page that makes money or makes contact: the booking form, the quote request, the checkout. For the diagnostic centre we work with in Kasur, that would be the MRI appointment page, not the homepage, because a homepage down for twenty minutes is embarrassing while a booking form down for a weekend is revenue the business never learns it lost.
Third, a keyword monitor on the page that must render data. A status 200 means the server answered; it says nothing about what the page shows. A keyword check that requires the string your booking calendar prints catches the case where the form loads and its data does not.
Fourth, the SSL certificate expiry, which most free plans watch alongside uptime. The padlock checks we described in an earlier article catch the error visitors see; expiry monitoring warns you weeks before it exists.
The self-hosted option, and when we pick it
Uptime Kuma is a self-hosted monitor with an MIT licence and 92,000 stars on GitHub, and it runs as one container on a server you already pay for. We pick it when the site is on infrastructure we manage, because the data stays with us and there is no per-minute cost. We do not pick it for the sites we host for clients, for one reason: if the monitoring runs on the same box as the site, the box going down takes the alarm with it. UptimeRobot’s own failures are rechecked across multiple checker nodes before an incident opens, and that independence from your server is the point you are paying nothing for.
Self-hosting also adds a maintenance task. The container needs updating, and the update schedule needs a monitor of its own. For a business with no technical staff, the hosted free plan is the better trade: less control, nothing to maintain.
The three mistakes that make monitoring pointless
The first is alerting to an inbox nobody reads. An email that arrives during a weekend shift and sits until Monday turned a five-minute outage into a three-day one. Every alert path we set up ends on a phone: push notification at minimum, SMS for the pages that take bookings.
The second is monitoring one URL and calling the site covered. The homepage and the booking form are different applications behind the same domain, and they fail independently. Two or three URLs, chosen by what the business loses when each one breaks, is the real coverage.
The third is paging on a single missed response. Networks drop single packets; a monitor that wakes you because one request timed out trains you to ignore it. UptimeRobot rechecks a failure from more than one location before declaring an incident, and any self-hosted setup should be configured the same way. And remember what our sweep showed: a fast, healthy-looking 404 is still an outage to a visitor, and a 200 with an empty data area is worse. Keyword checks on your key pages cover both.
FAQ
What is website uptime monitoring?
A service that requests your URLs on a fixed schedule, records the status codes and response times, and alerts you when they go wrong. A free plan watching five URLs at five-minute intervals will detect any outage that lasts longer than a coffee break. Self-hosted tools such as Uptime Kuma do the same on your own server.
How often should my site be checked?
Every five minutes is enough for a small business site; the free tier of UptimeRobot works at that interval and paid tiers drop to 60 seconds. Anything above ten minutes means an outage can hold through lunch and cost you the afternoon’s enquiries. Below one minute you mostly learn how fast your server is, not whether it is broken.
Can a website be up but still broken?
Yes, and our own pixel tracker proved it: track.dev2production.com answered with a valid TLS connection and a 404 on its homepage. Status codes alone cannot tell a working page from an error page that returns 200, so add keyword checks on the pages that carry data.
What does downtime actually cost in search rankings?
Google removes 404 pages from its index and drops pages that persistently return 5xx errors, per its crawler documentation. A one-hour outage is a hiccup; days of downtime unrank pages you spent months building. Monitoring shortens the outage, which protects the rankings already earned.
Set up website uptime monitoring today
Pick the three URLs a customer cannot do without: homepage, contact or booking page, and your most-visited service page. Add them to a free monitor at five-minute intervals, point the alerts at a phone, and trigger one test alert so you know the path works. Then run the check we run before calling a build done: request each URL from outside your network and read the status code, the first byte time and the body, because that is what your visitors and Google both see.
If you would rather hand this off, our web development services include building and monitoring sites with this sweep baked in, and the contact page reaches us directly. For the two failures that start before uptime is even the question, read our pieces on website backups for small business and checking a site’s padlock and certificate.