What One Hour of E-commerce Downtime Costs (and How to Justify It to the Client)

Search for "how much does downtime cost" and you always find the same numbers: thousands of dollars a minute, millions a year. The most-quoted figure by far — $5,600 a minute, or about $336,000 an hour — is attributed to a Gartner study from 2014: it's more than ten years old, it's an average calculated on large companies, and those who cite it report that Gartner itself presented it as the average of a very wide range.

If you look after online stores that make a few thousand euros a month, that figure is no use to you. Bringing it to a meeting is worse than useless: the client knows perfectly well that they don't lose $336,000 an hour, and the moment they realise it they stop believing the rest too.

Here we do the opposite sum: the minimum maths to estimate the real case of a small store, with every step in plain view, so you can redo it with your client's numbers instead of using ours.

The minimum formula

Three steps, none of them complicated.

1. Revenue per "trading" hour. Don't divide revenue by all the hours in the month: orders don't come in at four in the morning. Divide by the hours in which the store actually takes orders.

hourly revenue = monthly revenue ÷ (trading hours per day × days)

2. The time-slot concentration index. An hour on Saturday afternoon isn't worth the same as an hour on Tuesday at 10 a.m. If you have access to Analytics or the e-commerce back end, the number takes five minutes to pull out:

index = (share of revenue in that time slot)
        ÷ (share of hours in that time slot)

If the "Saturday 2-8 p.m." slot accounts for 20% of monthly revenue but is only 10% of trading hours, the index is 2: those hours are worth twice the average.

3. The estimated direct cost.

direct cost ≈ hourly revenue × index × hours of downtime

Then there are the indirect costs, which have to be added separately and estimated on their own — we'll get to them shortly.

The full example, with every assumption in plain view

Let's take a realistic online store for a small Italian agency. All the numbers below are stated assumptions, not measurements: they're there to show the method, and should be replaced with the client's own.

Starting assumptions:

• Revenue: €8,000/month
• Orders concentrated between 8 a.m. and 10 p.m. → 14 trading hours a day, for 30 days = 420 trading hours a month
• "Saturday afternoon" slot: concentration index of 2

The calculation:

€8,000 ÷ 420 hours = €19.05 per trading hour
€19.05 × 2 (Saturday afternoon index) = €38.10 per hour

An hour of downtime on a Saturday afternoon costs this store about €38 in missed revenue.

It's worth noting what happens if you get the denominator wrong: using the 720 hours of a whole month instead of the 420 trading hours, the same store would come out at €11.11 an hour — an underestimate of more than 40% compared with the real average for the hours in which it actually sells.

For reference, the same calculation on other sizes (still 420 trading hours a month, before the concentration index):

A necessary honesty: revenue is not loss

Those €38 are revenue missed in that hour: it's the highest defensible number, not the true one. Some of the orders aren't lost, they're postponed — people who wanted to buy come back the next day, and nobody knows how many. And the real economic damage isn't the revenue, it's the margin on those orders.

The real loss therefore lies between zero and €38, closer to the low end for a short outage in a weak time slot, closer to the high end for a long outage in a strong one. Telling the client this before they tell you doesn't weaken your case: it makes yours the only credible one of the day.

The costs that don't show up on an invoice

The agency's time. If an incident costs you an hour and a half between diagnosis, fix and communication, and your hourly rate is €50, that's €75 — which you absorb under a flat fee, not the client. Replace the two assumptions with your own and you have your number.

The store's customer service. Someone answers the people who write "the site isn't working." That's a person's time that the client pays for, and in long outages it isn't small.

Visitors' trust. There are no reliable numbers here, and whoever publishes some almost never explains how they measured them. The only thing you can say without making things up is qualitative: someone who finds a small, little-known store unreachable usually has no particular reason to try again later. It's an observation, not a statistic, and should be presented as such.

The number that really matters isn't "one hour"

Here's the point that changes the conversation. Let's go back to the €8,000/month store and compare two incidents identical in cause, different only in when they were discovered.

Case A — outage detected immediately. The site goes down on Saturday at 3:00 p.m., an automatic alert arrives within a few minutes, the problem is fixed in an hour.

1 hour × €38.10/hour = €38

Case B — silent outage. Same fault, same Saturday at 3:00 p.m., but nobody notices: the client writes on Monday morning at 9:00 a.m. We count only the trading hours lost (8 a.m.-10 p.m.):

Saturday 3:00-10:00 p.m. =  7 hours
Sunday 8:00 a.m.-10:00 p.m. = 14 hours
Monday 8:00-9:00 a.m.    =  1 hour
total                    = 22 trading hours

22 hours × €19.05/hour = €419

And that's a conservative €419, because it applies the average index to the whole period. Applying an index of 2 to just the 7 hours of Saturday afternoon, the figure rises to €552.

The same fault costs eleven times as much just because nobody noticed. (What to do in the minutes after you do notice is a separate checklist: the first 15 minutes of a site going down.) It isn't the outage that does the damage: it's the detection delay, and it's the only one of the two variables an agency actually controls. The host going down isn't up to you. The forty-two hours between the fault and the Monday phone call are.

Put next to the price: a monitoring item at €15/month makes €180 a year. A single silent weekend like the one in case B costs that store more than two years of that item.

How to use this calculation with a client, without selling fear

The wrong way is to arrive with the number already prepared and dramatic. The way that works is slower and holds up much better:

1. Ask for their numbers, don't use yours. "How much do you make in an average month? At what times do orders come in?" A calculation done with their data is theirs, and nobody argues with their own arithmetic.
2. Do the calculation in front of them, with the steps visible. It's three divisions: you can do them on a sheet of paper in two minutes.
3. State the limits before they do. "This is revenue, not margin, and some of the orders are just postponed." Someone who admits the limits of their own calculation is believed on the rest.
4. Shift the conversation from the outage to the delay. The right question isn't "how much does an hour of downtime cost you" — it's "how many hours would go by, today, before anyone noticed on a Saturday afternoon?" Almost always the client answers on their own, and the answer is your quote.
5. Close on proportion, not on fear. The ratio between the cost of monitoring and case B speaks for itself: said calmly, it's more convincing than said with alarm.

The calculation from the other side: the agency's

The same calculation, done on your price list instead of on their revenue, is what makes monitoring a sustainable line item and not a favour included in the fee: twelve clients at €15/month on the monitoring item make €180 a month, against the €59 of the Agency plan of Statuo, which covers up to 75 sites — €121 of margin every month, with no work beyond the initial setup of each site. (If you're also weighing other routes, the calculation against alternative tools is in the comparison index, self-hosting included.)

From the client's side, what they're buying is the hours of case B that won't happen: continuous checks from multiple independent locations, an outage declared only when at least two locations confirm it (so no false alarm ends up in their inbox), an alert by email or Telegram, and a status page with your branding that they can open whenever they like. The trial lasts 14 days, no credit card required.

A note on the numbers in this article

So that nobody has to take our word for it:

• The example figures (€8,000/month, 14 trading hours a day, index 2, €50/hour rate, 1.5 hours of intervention) are stated assumptions, not measurements: they're there to show the method. Every result derived from them comes from the divisions written out in plain view above, and can be redone in thirty seconds with different numbers.
• No percentages on lost customers, unrecovered orders or reputational damage: there's no reliable data on small Italian e-commerce stores and we didn't make any up. Where the data doesn't exist, the article says so.
• The Gartner figure ($5,600/minute, 2014) is cited only as a reference for what not to use with a small client, with the year stated explicitly so its age can be judged.
• Statuo's prices (€59/month for the Agency plan, up to 75 sites) are the public list prices; the €15/month for the resale item is the €10-20 range typically charged by agencies, not a price we impose.