A website going completely offline is the rare failure, and it is the one everybody notices within the hour. The expensive failures are the quiet ones, where the site keeps loading and something behind it has stopped working.
Every item below is something that leaves the site looking entirely normal to the person who owns it. That is precisely why they run for weeks.
The pattern
The costly failures do not look like failures.
When a site is down, the phone rings. When a contact form silently stops delivering, nothing happens at all — and nothing continuing to happen is indistinguishable from a quiet month. The absence of enquiries is not an alarm, which is what makes it the most expensive category of breakage there is.
The pattern repeats across every item in this list. The failure is invisible from the front page, there is no error for anyone to see, and the only signal is an absence: fewer leads, no data, a warning a visitor saw and you did not.
This is also why 'the site is fine, I checked it this morning' is not evidence of much. Loading the homepage confirms the homepage loads.
The list
Five things that break without telling you.
**The certificate expires.** Certificates have a fixed lifetime and renewal is usually automatic — until the automation breaks. When it does, every visitor gets a full-page browser warning, and the people who see it are the ones least likely to tell you: strangers, who simply leave.
**The form posts nowhere.** Forms fail in a specific and cruel way — the visitor still sees the thank-you message, because that is drawn by the page and not by the server. A changed mailbox password, a full inbox, a spam filter that started quarantining your own notifications, a plugin update: any of them ends delivery while the form keeps saying it worked.
**Tracking stops.** A tag removed during a redesign, a consent banner that now blocks it, a container that was never republished. Analytics does not warn you that a number is wrong; it shows a smaller number, and a smaller number looks like a slower month.
**A page falls out of the index.** A noindex that shipped from staging, a robots rule added during a migration, a canonical hard-coded by a template. Traffic to that page declines gradually rather than stopping, which reads as seasonality rather than a fault.
**DNS drifts.** A record edited for one purpose that broke another — most often mail. The website carries on resolving perfectly, which is exactly why nobody connects the two.
Why
Nobody owns checking, so nobody checks.
Most small business sites are built by one party, hosted by another and used daily by a third, and none of them has agreed to watch it. The developer's work ended at launch, the host monitors its own servers rather than your forms, and the owner is not looking for problems on a site that appears to work.
So these faults are usually found by accident — a customer mentioning they never heard back, a bank asking why the invoice email bounced, an agency noticing traffic fell off a cliff two months ago. By then the cost is already spent.
The fix is not complicated and it is not expensive. It is having a defined list of things that get checked on a schedule, and one person whose job it is.
Scope
What an external check can and cannot tell you.
An external scan reads what is publicly observable: the response and its headers, the certificate and its remaining validity, the presence of tracking code, indexability signals, and whether a form exists on the page with the fields it needs. That is a genuinely useful sweep and it takes seconds.
What it cannot do is prove delivery. No external tool can confirm that submitting your form results in an email arriving in your inbox, because doing so would mean submitting the form. The same applies to tracking: the presence of a tag in the source is not proof that events are firing.
So the honest division is this. Automate the observable part and run it often. Then test the unobservable part — the form, the notification, the reply path — yourself, by hand, on a schedule, because nothing else can.
- Certificate expiry and renewal, checked well before the date
- HTTPS reachable with no warning and no redirect chain
- Tracking tag present, and events actually arriving in the analytics property
- Indexability signals on the pages that matter
- DNS and mail records unchanged since you last agreed them
Doing it
A routine that costs nothing but a calendar entry.
Monthly, run an external scan of the site and read what changed since last time. Differences matter more than absolute scores: a category that was fine and is now not is worth more attention than a number that has always been middling.
Monthly, also submit your own contact form with a real message and confirm the email arrives where it should — in the inbox, not in spam. This is the single highest-value check on the list and it is the one no tool can do for you.
Quarterly, open your analytics and confirm the shape of the data still makes sense, and re-read your DNS and mail records against what you expect them to say. Everything here fits in under an hour a month, which is considerably less than the time spent working out why a quiet quarter was quiet.
Keep reading