What an hour of downtime actually costs you

Everyone agrees that downtime is bad. Almost nobody has a number, which is exactly why it never becomes a priority: you cannot budget for an anxiety. Here is how to get a defensible figure in five minutes, including for a site that sells nothing online.

The calculation, according to what your site does

"Useful hours" is the key, and it is where most calculations go wrong. Dividing by the 8,760 hours in a year gives a reassuring, false figure: nobody buys at four in the morning on a Tuesday. Take the hours during which your site is genuinely being used, and the number triples.

Three cases worked through

A shop turning over €200,000 a year. Sales concentrate into roughly 3,000 useful hours, so about €65 for each hour offline. Modest, until you notice that the outages nobody detects tend to last several hours, and that they often fall in the evening or at the weekend, which are the densest hours.

A tradesman receiving twenty enquiries a month. One enquiry in five becomes a job, a job is worth €3,000. Each enquiry is therefore worth €600 in expectation. Across 200 useful hours a month, an hour of downtime costs about €60, and a full day costs one enquiry, which is one job every five incidents.

An online service at €8,000 of monthly revenue. The direct loss from one hour is negligible, because subscriptions do not stop. The real cost is elsewhere: two visible outages in the same quarter, and churn moves. Here the figure to watch is not the hour lost, it is the number of incidents your customers saw.

The site that sells nothing online

This is the most common case, and the one articles on the subject systematically skip, because no formula can be written for it.

The loss is real all the same, and it takes three forms. A visitor looking for your opening hours who finds an error calls a competitor. An existing customer who cannot find your contact page writes to you on social media, or does not write at all. And an outage visible during a campaign, a trade show or a radio slot ruins precisely the one thing that was meant to pay.

The right measure here is not euros per hour, it is missed occasions: how many times a month does someone need this site, and what do they do when they cannot find it? Ten a week that each go elsewhere is a real number, and it is usually larger than the euros-per-hour figure a shop of the same size would produce. It is simply harder to put in a spreadsheet, which is why nobody does it.

The invisible part, usually larger

The interruption. Whoever fixes it drops everything. An hour of outage is rarely an hour of work: it is an afternoon broken in two, and the work that was meant to happen slides by a day.

The calls. Every customer who could not order and cared enough to tell you represents several who left silently. The time spent answering and reassuring counts too.

The search engine. A crawler that lands on an error page comes back less often. Nothing dramatic in one incident; measurable across several, and it is an effect that recovers slowly.

The trust. This is the part you cannot invoice. A visitor who finds a dead site does not come back to check whether it has been fixed. They are already on the next result, and they will never know you repaired it in ten minutes.

What costs almost nothing, despite appearances

An article that inflates everything loses its credibility with the first reader who actually counts. Three things cost less than people say:

• A night-time outage on a local business site, when nobody browses at those hours. It is unpleasant and it does not add up to a figure.
• An outage of a few minutes, if it repairs itself. The visitor reloads, and most of them notice nothing.
• An announced outage, during maintenance planned for a quiet hour. This is the case where a 503 is the correct answer, and it exists for exactly that.

The two most common calculation errors

Dividing by a whole year. This is the error that makes the figure harmless. A business site does not work 8,760 hours a year, it works two or three thousand, and an outage does not fall at random: it falls while the server is being used, which means during the hours that count. The per-useful-hour figure is the only one describing what actually happens.

Counting revenue instead of margin. A shop losing €200 of sales does not lose €200: it loses the margin, and it keeps the stock. The gap matters on low-margin businesses, and it matters less than people think, because fixed costs keep running throughout the outage.

Of the two, the second is venial. The first distorts everything, and it explains why so many people conclude that downtime "does not cost much" before discovering otherwise on a Friday evening.

What this figure is actually for

It is not there to frighten you. It is there to settle three concrete trade-offs: how much you accept to pay for more reliable hosting, how often you check, and who gets woken up at night. Without a figure, those three questions are decided by feel, and feel always says "later".

The calculation that actually helps

Do not look for a perfect figure. Answer one question instead:

If my site had been down for three hours yesterday afternoon, what would I have lost, and how long would it have taken me to find out?

The second half of the question is the one that matters. Most businesses cannot reduce the number of outages: they depend on a host, a plugin, a certificate. What they can reduce is the time between the outage and the moment they hear about it.

And that delay is entirely in your hands, unlike everything else: your host will not tell you, since from their point of view the machine is answering.

Where the money actually is

An outage detected in four minutes costs the fix. The same outage discovered the next morning by a customer costs the fix, the customer, and the conversation that follows.

The technical difference between the two is small: a check that runs by itself and an email. The financial difference is the whole point. It also settles the right setting: the checking interval is chosen from this figure, not from what a tool happens to offer.

One last use for this calculation, and not the least: it goes into a report. A client reading "99.95% availability" understands nothing; a client reading "two incidents, forty minutes in total, no orders lost" understands immediately, and that is what a monthly report should contain.

For what comes immediately next: the first fifteen minutes when a site stops responding.