Setting Up Alerts Instead of Reports

Pull against push

A report requires somebody to open it, on a schedule, and notice something.

An alert arrives when a condition is met, whether or not anybody was looking.

Most of what a small business needs from measurement is the second kind, because the useful cases are all failures rather than trends.

And a failure discovered on the fifteenth of the following month has already cost a month.

The four worth having

• The site is down.
• No form submissions in several days.
• Traffic has fallen sharply.
• The certificate is about to expire.

The second is the one almost nobody sets and it catches the most expensive failure a small business site has. A form that stops delivering is silent, and the usual discovery is a customer asking why nobody replied.

The submissions alert is the important one

Worth setting up even if you do nothing else here.

The condition is simple: no submissions recorded in a period longer than your normal gap.

For a business receiving eleven a month, three days with none is unusual enough to check and common enough not to be annoying.

For one receiving two a month, a fortnight is the right threshold.

Set it from your own history rather than a default, since the whole value is in it firing rarely and meaning something when it does.

Most analytics tools can send this, and where yours cannot, a monthly reminder to submit your own test form covers most of it.

A worked example

A business had a form stop delivering after a plugin update on a Thursday.

They found out eleven days later when a customer telephoned to ask whether they had received her message.

Three enquiries had been lost and one of them was a substantial job.

They set an alert for three days with no submissions, which has since fired twice: once for a genuine break and once during a quiet week over Christmas.

The false one took ten seconds to dismiss.

The genuine one was caught the same day.

Tune it or you will ignore it

Which is the failure mode of all alerting.

An alert that fires weekly for no reason gets filtered, and then the real one arrives in a folder nobody opens.

So set thresholds generously at first and tighten them only if something is being missed.

Four alerts that fire two or three times a year between them is the right frequency.

If any one of them is firing monthly, its threshold is wrong rather than your business being unstable.

Review them once a year and remove any that have never told you anything.

Send them somewhere they will be seen

Since delivery decides whether it works.

To a person rather than to a shared address nobody monitors.

To a phone rather than to an email account only checked at a desk, for the ones that need same-day action.

And to two people where the business has them, since one person is on holiday at some point.

An alert delivered to an unmonitored address is the same as no alert while feeling like coverage, which is worse than knowing you have none.

Some alerts belong outside analytics

Worth separating, since the most valuable ones are not analytics at all.

Your domain expiring, your hosting renewal, and a payment card about to expire on a subscription are all date-based and all cause outright failures.

Each of those is a calendar entry rather than a monitoring tool, which makes them free and reliable.

Put the renewal dates in the same calendar as everything else, with a reminder a month ahead rather than a week.

A domain that lapses because a notice went to an address nobody reads is the most avoidable outage a business can have, and it happens constantly.

What not to alert on

Since the temptation is to set twelve.

Ranking changes, which move constantly and require no same-day action.

Traffic rising, which is never urgent.

Individual page changes, which produce noise.

And anything you would look at rather than act on, which is the same test as everywhere else.

An alert should mean something is broken, not that something is interesting.

That single test keeps the list at four rather than twelve, which is what keeps them being read.

They replace most of the report

Which is the point of the exercise.

Once four alerts are covering the failures, the monthly report has only one job: the trend in the number that matters.

That is a single figure and a comparison, which takes two minutes rather than twenty.

Everything the longer report was watching for is now watching itself.

This is the arrangement most small businesses should have arrived at years ago and almost none have.

Write down what each one means

Since an alert arriving at eleven at night has to be actionable by whoever receives it.

One line per alert: what it means, how urgent it is, and the first thing to check.

Otherwise the person on call receives a message saying a threshold was crossed and has no idea whether that requires waking somebody or waiting until morning.

Keep it with the hosting details, where somebody looking for a login will find it anyway.

That also makes the alerts transferable, which matters when the person who set them up is the one on holiday.

The counter-case

Alerts need maintenance.

They break silently when a property changes, an address stops working, or a tool is replaced, and a broken alert is indistinguishable from a quiet period.

Test each one once a year by deliberately triggering it, which is the only way to know it still works.

And for a business with genuinely stable patterns, a monthly glance may catch everything an alert would.

Set four alerts, tune them to fire rarely, send them to a person's phone, and cut the report to one figure.

The setup

1. Set an uptime alert.
2. Set a no-submissions alert.
3. Set a sharp traffic drop.
4. Set a certificate expiry.
5. Send them to a phone.
6. Set thresholds generously.
7. Test them annually.

Step two is the one that pays for the whole exercise, since a form that stops delivering is the most expensive silent failure a small business site has.

The site-availability version is covered in uptime monitoring that tells you first.

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.