Downtime cost calculator
Outages are argued about in minutes and budgeted in money, and the conversion is what usually goes missing. This estimates what an hour offline costs you across the three components that are actually measurable. It runs entirely in your browser.
Your numbers
Per minute: . Currency is whichever one you entered - The arithmetic does not care.
Worked examples
The inputs below are assumptions, not reported figures. We do not know what any outage actually cost these companies, and neither does anyone outside them. What the examples show is the shape of the arithmetic for three very different businesses. Press a button to load the numbers, then replace them with your own.
An AI assistant, mid-morning
The interesting case, because the damage is not confined to the provider. A growing share of ordinary software calls a model API in the background, so when one degrades, support widgets stop replying and summaries never appear in products with no visible AI feature at all. We monitor Claude and ChatGPT every 20 minutes, and their recorded history is on those pages. Capacity, not crashes, is the usual cause: rate limits tighten and queues lengthen rather than anything returning an error.
- Monthly online revenue1,200,000
- Gross margin65%
- Staff blocked18
- Length1.5 hours, peak
An online shop on its busiest day
The most concentrated cost in this directory. Browsing survives loads that checkout cannot, so the first symptom is a basket that will not proceed while every product page loads instantly. An hour lost during a sale is not an average hour: The multiplier is doing most of the work in this calculation, which is why it is a separate field.
- Monthly online revenue400,000
- Gross margin35%
- Staff blocked6
- Length45 minutes, campaign peak
An internal tool with no revenue
Worth running because the answer is often uncomfortable. Nothing is sold through it, so the revenue line is zero and the entire cost is people who cannot do their job plus the work of catching up afterwards. This is the case where a per-minute figure most often changes somebody’s mind about a hosting bill.
- Monthly online revenue0
- Gross marginn/a
- Staff blocked40
- Length3 hours, working day
Every one of these services has outages, including the two named above. What we can show you is the measured record: every outage our checks have confirmed, with real durations, and how long they typically last by category.
What this includes, and what it deliberately leaves out
Most downtime cost calculators produce a large number by adding speculative categories together. This one only counts the three you can defend in a meeting, and it uses gross profit rather than revenue, because an order you did not fulfil also did not cost you anything to fulfil.
Lost gross profit
Your monthly revenue, reduced to an hourly figure, multiplied by your margin and by the traffic multiplier for the time of day. The multiplier matters more than anything else here: An hour down at 3am and an hour down during a campaign are different events by an order of magnitude, and averaging them is how outages get under-priced.
Idle staff time
People who cannot do their job while the system is down, at their fully loaded hourly cost. Use the number genuinely blocked, not the headcount - And note this cost is real even for an internal tool with no revenue attached to it at all.
Recovery work
Estimated at 1.5× the outage length in engineering time. Incidents are not over when the service comes back: There is the backlog to clear, the data to reconcile, the customers to answer and the write-up to do. This is the component people forget, and on a short outage it is frequently the largest.
What is not counted
Churn, reputational damage, SLA credits owed to your own customers, and the orders that go to a competitor and never come back. These are real and often dominate the true figure, but any number put on them is invented, and an invented number is what gets the whole estimate dismissed. Treat the total here as a floor rather than an estimate.
The number that usually settles the argument
Per-minute cost is the figure worth carrying out of this page. It converts every subsequent decision into the same unit: If an outage costs you $40 a minute, then twenty minutes of faster detection is worth $800 an incident, and the monitoring that provides it is cheaper than that by a wide margin. Availability spending is nearly always argued about as a cost, and nearly always resolved once somebody produces a per-minute figure.
To turn an availability promise into the hours it actually permits, pair this with the uptime calculator; to see what real incidents have cost in duration rather than money, our outage log records every one we have measured. It is also the honest way to find out that you should not spend more. Plenty of services genuinely cost very little per minute offline, and the correct response to that is to accept the risk and spend the money somewhere else. A calculator that only ever justifies buying more is not a calculator.