Uptime Monitoring That Tells You First

Finding out from a customer

The usual way a small business learns its site is down is a message from somebody trying to use it.

That is slow, since it depends on somebody both noticing and bothering to tell you, and it is a poor impression at exactly the moment you would rather make a good one.

Monitoring removes both problems for very little money, and free tiers cover what most small businesses need.

The reason it is worth writing about is not whether to have it. It is that most of what gets set up would not catch the failures that actually happen.

The homepage is the wrong thing to watch

The default setup, and the one with the largest blind spot.

A homepage check confirms that the server is responding and a page is being returned. It says nothing about whether anything works.

A site can serve a perfect homepage while the contact form silently fails, the checkout errors, the database is unreachable on interior pages, or every product page returns nothing.

Those are the failures that cost money, and a homepage monitor reports everything as fine throughout.

Monitor two or three additional addresses: a service page, the contact page, and for a shop, a product page and the basket.

What to monitor

• The homepage, as a baseline.
• An interior page that uses the database.
• The contact or booking page.
• The certificate, with a warning before expiry.
• The domain expiry date.
• Page content, not only that a page loaded.

The last is the one that separates useful monitoring from the default. Checking that a page contains an expected word catches a site returning an error page with a successful status, which a simple check reports as working.

Alerts have to reach a person

The failure that makes monitoring pointless, and it is extremely common.

An alert sent to a general inbox, or to an address checked once a day, tells you about an outage after it ended.

Send it to a phone: a text message, a push notification, or a messaging application somebody actually watches.

Send it to two people rather than one, since the person who normally deals with it is the person who will be driving.

And agree what the recipient does, which for most small businesses is telephoning one named person rather than attempting a fix.

Monitoring that nobody acts on is a log rather than an alarm.

Tune it or you will ignore it

The practical problem that ends most monitoring arrangements within a month.

A check running every minute, alerting on a single failure, will produce false alarms from ordinary network noise, and after four of those at unsociable hours somebody switches the notifications off.

Require two or three consecutive failures before alerting, and check from more than one location if the tool allows it.

Set a sensible interval: every five minutes is ample for a small business site, and every minute mostly produces noise.

And configure a recovery notification, so you know it came back without having to check.

Monitoring that cries wolf gets disabled, and a disabled monitor is worse than none because you believe you are covered.

A worked example

A business had monitoring on their homepage and considered themselves covered.

Their contact form stopped delivering after a plugin update and nobody knew for nine days, because the homepage was fine throughout and the monitor never fired.

They found out when a customer telephoned to ask why nobody had replied to an enquiry sent the previous week.

They added three checks: an interior page, the contact page, and a content check confirming the homepage still contained their phone number.

They also added a monthly reminder to submit the form themselves, since no external monitor can confirm that a message actually arrived in a mailbox.

The manual check is the one that would have caught it, which was the uncomfortable finding.

What monitoring cannot see

Worth being clear about, so it is not over-trusted.

It cannot tell you that a form submitted successfully and the email never arrived, since the page responded correctly.

It cannot tell you that payments are failing at the processor, or that a booking system is accepting appointments into a calendar nobody reads.

And it cannot tell you the site is slow enough to be losing people, unless you have configured a response time threshold.

Those need a person doing a periodic manual check: submit the form, place a test order, make a test booking.

Monthly is enough, and it catches the category of failure that produces no error at all.

Keep the history

A secondary benefit that turns out to be the more useful one.

A monitoring tool records every incident with times, which is exactly the record worth having when judging a host across a year.

It also gives you an independent account, rather than relying on a status page written by the party responsible.

Export or note the summary every few months, since free tiers frequently retain only recent history.

That record is what makes a conversation with a host specific, and what tells you whether things are getting better or worse.

The counter-case

Monitoring can be more than a small site warrants.

For a brochure site with a handful of enquiries a month, knowing about a forty-minute overnight outage changes nothing, because nobody was going to act until morning regardless.

There is also a cost in attention: alerts at unsociable hours for something nobody can fix until Monday produce anxiety rather than uptime.

The proportionate arrangement for a small brochure site is a single free check with alerts by email, which tells you the next morning and costs nothing.

Full alerting to a phone belongs to sites that take orders or bookings, where an hour genuinely costs something.

Setting it up

1. Monitor three or four addresses, not just the homepage.
2. Add a content check, not only a status check.
3. Set the interval to five minutes.
4. Require two failures before alerting.
5. Send alerts to a phone, and to two people.
6. Agree what the recipient does.
7. Submit your own form monthly, manually.

Step seven catches what no monitor can, and it is the check that would have prevented the example above.

Preparing a site to run unattended is covered in a website that runs without you for two weeks.

Frequently asked questions

West Coast Media Solutions Inc. provides web design, web development, hosting, digital marketing, and business consulting to organisations across Canada, drawing on more than twenty-five years in the field.