Free tool · Model v1.0
What does an hour offline actually cost you?
Most answers to this question are a recycled statistic from 2014 multiplied by your revenue. This one asks for your numbers, shows every step of the arithmetic, and tells you what it refuses to estimate.
Why another calculator
The figure most sites quote, $5,600 a minute, comes from a Gartner study published in 2014. It was a cross-industry average across large enterprises, Gartner flagged it as an average, and it has not been meaningfully refreshed since.
Other estimates from the same period range from roughly $2,300 to $9,000 a minute. That spread is the point: an average across industries tells your business nothing useful.
The calculator
Your numbers, not an industry average
It starts with example figures so you can see how it behaves. Replace them with your own. Nothing is sent anywhere, and the result updates as you type.
4.0 hours offline, peak trading hours
£3,092
£773 for every hour the site was down.
Where that comes from
Gross profit not earned£237
£1,314 of revenue blocked, less 40% recovered later, at 30% margin
Revenue you never take is not the same as money lost: you avoid the cost of the goods too, and some customers come back. This line is the profit that genuinely does not arrive.
Value of enquiries never made£1,402
438 missed visits, 3.5 lost enquiries at £400 each
Visitors who arrive at a broken site and do not return. The recovery rate matters most here: people researching a considered purchase are more likely to try again than an impulse buyer.
Advertising spent on a broken site£44
£5 an hour × 2× × 4 hours
Clicks are charged when they happen, not when the page loads. Pausing campaigns during an incident is one of the few things that reduces this in real time.
Your team's time£810
3 people × £45 an hour × 6 hours
Count the hours actually consumed, not the length of the outage. Incidents pull in people who were not fixing anything, and the rest of the day rarely recovers.
Emergency help£600
£150 an hour × 4 hours
Out-of-hours developer or agency rates, which are the rates you pay when you had no plan. Retained support is usually cheaper precisely because it is not an emergency.
Revenue blocked during the outage came to £1,314. Most calculators would stop there and call that your loss. We subtract the 40% who buy later and count gross profit rather than turnover, which is why the figure above is smaller and easier to defend.
What your uptime is worth per year
Trading losses only, at a normal hour. Staff time and emergency help are per incident, so leaving them out keeps this conservative.
| Uptime | Time offline | Yearly cost |
|---|---|---|
| 99.9% | 8.8 hours | £1,843 |
| 99.5% | 1.8 days | £9,216 |
| 99% | 3.7 days | £18,432 |
| 95% | 18.3 days | £92,160 |
The gap between the first two rows is what a maintained site is worth to you in a year. It is usually more than maintaining it costs.
What this deliberately leaves out
Search rankings
A short outage almost never costs rankings: search engines retry. A long or repeated one can, but the effect depends on crawl timing, the status code returned and how long recovery takes. Anyone who gives you a figure for this is guessing.
Reputation and churn
Real, and genuinely unmeasurable from the outside. It depends on what you sell, how often people buy, and whether a competitor is one click away.
Contractual and regulatory exposure
If you owe customers an uptime SLA, or operate under rules on availability or reporting, the cost of a breach is specific to your contracts. Read them rather than estimating.
The work that did not happen
The project that slipped because the team spent Thursday firefighting. It is a genuine cost and nobody can price it credibly, so we leave it out rather than inflate the total.
The method
Two decisions that make this number smaller
Almost every downtime calculator multiplies hourly revenue by the length of the outage. That produces a large, quotable figure and an indefensible one.
Revenue is not profit
A sale you never make also means goods you never ship and card fees you never pay. The cost to the business is the gross profit that did not arrive, not the turnover that did not pass through. We ask for your margin and use it.
Not every customer is gone
Some people come back an hour later and buy anyway. How many depends on what you sell and whether a competitor is one click away. There is no reliable public figure for this, so it is an input you control rather than a constant we picked. It also has the widest range of anything in the model: at 0% the trading losses count in full, and at 100% they disappear entirely. If you have figures from an incident of your own, use those.
Hours are not interchangeable
An outage at 3am and an outage during a campaign are not the same event. The multipliers are rules of thumb rather than findings, which is why they are a visible choice and not buried in the formula.
Check the arithmetic
Every line in the result shows the sum behind it, so a finance reviewer can follow it without taking our word for anything.
- A month is 730.5 hours, not 720. Using 30 days understates hourly figures by about 1.5%.
- The yearly table counts trading losses only, because staff time and emergency help are per incident and do not scale with hours.
- Four categories of cost are named and excluded rather than guessed at, so the total is a floor and not a sales figure.
Use it on your own site
Embed the calculator
Paste this where you want it to appear. It is free to use, including commercially, as long as the attribution link stays visible.
About the model
What people ask before putting this figure in a business case.
Why is your number lower than other downtime calculators?
Because most of them multiply your hourly revenue by the length of the outage and stop there. That assumes every blocked sale is lost forever and that revenue and profit are the same thing. Neither is true. This model subtracts the customers who come back and counts gross profit rather than turnover, which produces a smaller but defensible figure.
What about the famous $5,600 per minute?
It comes from a Gartner study published in 2014 and has been repeated ever since. It was a cross-industry average across large enterprises, Gartner flagged it as an average, and it has not been meaningfully refreshed in over a decade. Other estimates from the same period range from roughly $2,300 to $9,000 a minute, and that spread is the point: an average across industries tells a specific business nothing. Analysts at Enterprise Management Associates now describe the figure as an urban legend.
Which assumption changes the answer most?
The two recovery rates: what share of blocked customers and enquirers come back later. They decide how much of the trading loss is permanent, and taking them from 0% to 100% moves the trading lines between their full value and nothing at all. No other input has that range. There is no reliable public data on them either, which is why they are inputs rather than constants we picked. If you have figures from a past incident of your own, use those instead.
Can we use this in a business case?
Yes, and that is what it is for. Every line shows its arithmetic so a finance reviewer can check it, and the exclusions are listed so nobody can accuse you of padding. Cite the version and link back.
Can we put the calculator on our own site?
Yes. There is an embed snippet further down the page. It is free to use, including commercially, as long as the attribution link stays visible.
Now work out whether it has to keep happening
If the yearly figure is bigger than you expected, the next question is why the site goes down at all. That is usually an operations problem with a short list of causes.