Downtime Cost Calculator: What an Hour Offline Costs You
Lost sales and idle staff, per hour and per minute, for one outage and for a year at the uptime you have now. The number to have in hand before deciding what to spend on avoiding it.
Your Business
Rough figures are fine. The result is only as exact as the guesses, but the order of magnitude is what matters.
Cost per staff hour includes employer's costs: salary times about 1.3, divided by about 1,650 working hours a year.
A Year at Each Uptime
| Uptime | Downtime a year | Cost a year |
|---|
What the Number Leaves Out
Customers who don't come back. A first-time buyer who meets an error page rarely tries again, and a SaaS customer who loses a morning starts reading competitors' pricing pages.
Contract penalties. If you promise customers an uptime figure, a long outage can mean service credits on top of the lost sales.
Lost data. An outage that ends in a restore from last night's backup also loses everything since. The RTO and RPO calculator shows how much.
The people fixing it. Hours of developer time pulled away from planned work, often late at night.
So treat the result as the least an outage costs, not the most.
Using the Number
Compare it with what prevention costs. If an hour down costs £500 and outages run to ten hours a year, £5,000 a year buys a lot of monitoring, a second server or someone looking after the setup. If an hour costs £20, a simpler setup and a good backup may be enough.
Shorter outages beat fewer outages. Most of the cost is in how long an outage lasts. Alerts within a minute, a runbook someone can follow at 2am, and restores that have been practised cut hours to minutes.
Most outages are avoidable. On small setups the usual causes are a failed deploy, an expired certificate, a full disk or a patch applied without a plan, not hardware failure.
Questions
How do you calculate the cost of downtime?
Add the revenue you lose for each hour the site or app is down to the cost of the staff who can't work. Lost revenue per hour is the yearly revenue that depends on the system, divided by the hours you trade in a year, less the share of sales that simply come back later. Staff cost is the number of people affected, times their hourly cost, times how much of their work stops.
What is the average cost of downtime?
There isn't a useful average. The often-quoted figure of $5,600 a minute comes from a 2014 Gartner estimate for large enterprises. For a small online business it may be tens or hundreds of pounds an hour. Work out your own: it is the number that justifies, or doesn't, spending on redundancy and monitoring.
Does downtime only cost lost sales?
No. Staff stand idle or work on the outage, customers ask for refunds or service credits, and some don't come back. Search engines may drop pages that fail when they crawl. The calculator counts sales and staff; the rest is real but hard to put a figure on, so treat the result as the minimum.
Do lost sales come back after an outage?
Some do. Customers of a shop they already trust often return later in the day; new visitors from an advert usually don't. The "sales that come back later" setting is your best guess at that share: 20 to 40% is common for established online shops, and close to zero for advertising-led traffic.
How much downtime does 99.9% uptime allow?
About 8 hours 45 minutes a year, or 43 minutes a month. Multiply that by your cost per hour to see what the allowed downtime costs you, and compare it with what moving to 99.95% or 99.99% would take. The uptime calculator has the figures for every SLA.
How can I reduce the cost of downtime?
Shorten outages, more than chasing perfect uptime. Monitoring that alerts someone within a minute, a written recovery runbook, and backups that have been restored all cut the hours. Most outages on small setups are self-inflicted: failed deploys, expired certificates, full disks and unplanned patching.
Want Fewer, Shorter Outages?
I look after servers for agencies and software teams: monitoring that wakes someone, patching in planned windows, and restores that have been tested.