Website downtime cost is not a single number, it is a formula: cost per minute = revenue lost per minute + payroll burn of everyone responding + recovery and cleanup + the slow costs (churn, refunds, SLA credits, reputation). For most small and mid-sized businesses, the direct revenue loss lands somewhere between $0.50 and $20 per minute, while the response payroll alone runs $2 to $8 per minute the moment engineers drop what they were doing.
That means a typical one-hour outage at a $2M-a-year online store costs roughly $228 in lost sales and several hundred more in salaried time, wasted ad clicks and support load. The rest of this page shows the arithmetic, the tables, and the parts of the bill people routinely forget.
The downtime cost formula
Four components. Three of them you can calculate today.
- 1. Revenue per minute. Annual online revenue ÷ 525,600 minutes in a year. This is the number everyone reaches for first, and for most small businesses it is the smallest line on the bill.
- 2. Response payroll. (Number of people responding) × (fully loaded hourly cost) × (hours the incident consumes). Fully loaded means salary plus taxes, benefits and overhead, usually 1.3 to 1.5 times base pay.
- 3. Recovery and cleanup. Data repair, replayed jobs, re-sent emails, the postmortem, the customer apologies, and the two days of planned work that got shoved into next sprint.
- 4. The slow costs. Churn, refunds, SLA credits, abandoned carts that never come back, and the trust you spend explaining yourself. Real, large, and impossible to invoice precisely.
Add the first three for a defensible floor. Treat the fourth as the reason the floor is an understatement, not a target.
Worked example 1: a $2M ecommerce store
An online store doing $2,000,000 a year divides that across 525,600 minutes:
- $2,000,000 ÷ 525,600 = $3.80 per minute of average revenue.
- A 60-minute outage therefore costs about $228 in direct revenue ($3.80 × 60).
- Two engineers and a founder spend 2 hours on it. At a fully loaded $100 an hour each, that is $600 of payroll.
- Direct total: roughly $828, before a single refund, ad click or support ticket.
Notice which line is bigger. The founder who assumed the outage cost "a few hundred in sales" was right about the sales and wrong about the bill. And the $228 is an average: outages cluster in peak hours because peak load is what breaks systems, so a checkout failure at 8pm on a Tuesday can easily be worth 3 to 5 times the average minute. Catching that specific failure fast is what ecommerce website monitoring is for: check the checkout path, not just the homepage.
Worked example 2: a $500K ARR B2B SaaS
Same division: $500,000 ÷ 525,600 = about $0.95 per minute. A full hour of downtime "costs" $57 in revenue. Nobody sensible believes that, and they are right not to.
Subscription revenue does not stop when the app goes down; it stops at renewal. The real exposure for a B2B SaaS is somewhere else entirely:
- SLA credits. If your contract promises 99.9% and you breach it, you owe credits on the monthly fee. A 10% credit across a $40,000 monthly book is $4,000 for one bad month, and the customers who claim it are usually the ones about to renew.
- Churn. One enterprise customer at $30,000 a year who leaves because of a visible outage costs 500 times more than the "revenue per minute" said the outage was worth.
- Sales pipeline. Prospects run trials during their evaluation window. A 502 during someone's proof of concept is a lost deal you never hear about.
Per-minute revenue is the wrong lens for subscription businesses. Price your downtime by contract exposure instead, then read the mechanics of 99.99 uptime meaning and SLA credits before you sign the next one.
Revenue lost per minute of downtime, by annual revenue
Straight division, 525,600 minutes in a year. Find your row. This is the direct revenue line only, not the full cost.
| Annual online revenue | Lost per minute | Lost per hour |
|---|---|---|
| $250,000 | $0.48 | $29 |
| $500,000 | $0.95 | $57 |
| $1,000,000 | $1.90 | $114 |
| $5,000,000 | $9.51 | $571 |
| $10,000,000 | $19.03 | $1,142 |
Two honest caveats. First, these are averages across every minute of the year, including 3am on a Sunday, so a peak-hour outage costs multiples of the row you are reading. Second, not all of that revenue is truly lost: some customers come back an hour later. A common working assumption is that 50% to 70% of interrupted transactions return for a short outage, and far fewer for a long one or a first-time visitor who has never heard of you before.
Uptime percentage to allowed downtime
Cost is minutes times money, so your uptime target is really a budget in minutes. Here is what each level buys you before the number is broken:
| Uptime | Downtime per year | Per month |
|---|---|---|
| 99% | 3d 15h 36m | 7h 18m |
| 99.9% | 8h 46m | 43m 50s |
| 99.99% | 52m 36s | 4m 23s |
| 99.999% | 5m 15s | 26s |
Multiply the year column by your per-minute number and you have your annual downtime bill at each target. For the $2M store above at $3.80 a minute: 99% uptime allows 5,256 minutes a year, which is nearly $20,000 of lost revenue as a matter of policy. At 99.9%, the same policy costs about $2,000. That gap is the actual business case for better hosting and better monitoring, and it is usually a much easier conversation than arguing about a good uptime percentage in the abstract.
alertping
The bill is measured in minutes. So is the fix.
AlertPing checks your site from 3 regions and calls, texts or Slacks a human the moment it fails. Plans from $19 a month, with 30-second checks from $59, which is less than most single outages cost in payroll alone.
The costs everyone forgets
The four lines below are where a "$228 outage" quietly turns into a four-figure one. None of them are speculative.
- Your ads keep spending while the page is broken. This is the sharpest one. Google Ads has no idea your landing page is returning a 502. If you spend $5,000 a month on paid search, the average hour buys about $7 of clicks, and a peak business hour buys several times that. Every one of those clicks lands on an error page, converts at zero, and still gets billed. You are paying full retail to show strangers your worst moment.
- Support load. An outage that touches 2% of a 10,000-customer base can generate 200 tickets. At 6 minutes each, that is 20 hours of support time, plus the backlog it creates for everyone who wrote in about something else.
- Abandoned carts and lost signups. A checkout error at the payment step does not just lose the sale, it loses the trust. Recovery emails claw some of it back; the rest is gone.
- SEO crawl impact. A short outage is harmless: Googlebot retries. A long one, or a pattern of repeated 5xx responses, slows crawl rate and can drop pages out of the index. This is not a per-minute cost, it is a delayed one that shows up in next month's traffic.
- Team trust. Every unplanned incident spends morale. Two 3am pages in a month and your best engineer starts reading recruiter emails. That cost never appears in a spreadsheet and dwarfs everything above it.
Incident payroll and infrastructure burn come out of the same run-cost budget as your servers and your tooling, and teams that start pricing downtime honestly usually run into the next question fast: whether they can even say what their cloud and SaaS spend actually is in a given month. Both questions have the same shape. You cannot manage a cost you have never measured.
About that "$5,600 per minute" figure
You will see a number like this cited everywhere, usually with no source, usually attributed to a survey that is over a decade old. Here is the plain truth: those averages come from samples of large enterprises, where a single minute of a trading platform or a global logistics system genuinely does cost thousands. Averaged across such a sample, the figure is arithmetically true and practically useless.
If you run a $500K SaaS or a $2M store, quoting an enterprise average to your own team is worse than quoting nothing, because it makes the whole exercise feel fictional. Nobody believes their homepage costs $5,600 a minute, so nobody believes the rest of the analysis either. Use the formula. It takes four minutes, it uses your own revenue, and you can defend every line of it in a budget meeting.
How much does website downtime cost per minute?
Divide your annual online revenue by 525,600. A business doing $1 million a year online loses about $1.90 per minute in direct revenue. Then add response payroll: three engineers at a fully loaded $100 an hour is $5 a minute on its own. Most small businesses land between $2 and $25 per minute in direct cost, with churn and reputation costing more over time.
How much does one hour of downtime cost?
For a business doing $1 million a year online, one hour costs about $114 in lost revenue, plus roughly $300 in payroll if three people spend the hour on it, plus wasted ad spend and support tickets. Call it $500 for an ordinary hour and 3 to 5 times that for a peak hour, because outages cluster where load is highest.
What is the average cost of IT downtime?
There is no average worth using. The widely quoted per-minute figures come from old surveys of large enterprises and say nothing about a business doing under $10 million a year. Calculate your own instead: revenue per minute, plus responder payroll, plus recovery, plus contract exposure. Your number will be far lower and far more useful.
How do you calculate the cost of downtime?
Use this formula: (annual revenue ÷ 525,600) × minutes of downtime, plus (responders × fully loaded hourly rate × hours), plus recovery work, plus SLA credits owed. Multiply by 3 to 5 if the outage hit peak hours. That gives you a defensible floor. Churn and reputation sit on top of it and are always larger than they look.
Cutting the cost means cutting the minutes
Every term in the formula is multiplied by minutes. You cannot change your revenue per minute this quarter, and you cannot change what your engineers cost. You can change how many minutes an outage runs, and the biggest slice of most outages is not the fix, it is the gap before anyone noticed.
Detection time is the one variable you can buy down cheaply. A 30-second check interval bounds the time an outage can hide at 30 seconds. A 5-minute interval means a failure can run for the better part of five minutes before your monitoring even looks, and if the result goes to an email nobody reads on a Saturday, add another hour. At $3.80 a minute plus payroll, shaving 20 minutes off a single incident pays for a year of monitoring several times over.
That is the whole argument for an external website monitoring tool: it watches from outside your infrastructure, so it still works when your infrastructure does not, and it turns detection from something you hope for into something with a number attached. AlertPing runs 30-second checks confirmed from three regions and sends downtime alerts by SMS, Slack, email and webhook, at flat uptime monitoring pricing from $19 a month. Whatever you use, price your own outage first. The number is usually bigger than the tool, and that is the point.