The math nobody wants to do
There’s a reason outage budgets get approved after an outage and not before: until you have lived through one, downtime feels free. The research says otherwise. ITIC’s hourly-cost survey found that for 91% of organizations, a single hour of downtime costs $300,000 or more — and for 41% of mid-size and large enterprises, somewhere between $1M and $5M per hour. Gartner’s long-standing baseline lands in the same neighborhood: roughly $5,600 per minute.
If those numbers sound enterprise-only, consider the small end: ITIC pegs even a company under 25 people at around $100,000 per hour of downtime — about $1,670 every minute your systems are dark.
Where the money actually goes
Lost revenue is only the visible part. The costs that really hurt are quieter:
- Payroll for idle staff. Forty people being paid to watch a spinner is real money, every hour.
- The trust bill. Splunk and Oxford Economics estimate that large companies lose an average of $300 million a year to unplanned outages — and that a company’s stock drops about 3.4% on average after a major incident. Your version of that is the customer who quietly moves to a competitor.
- The recovery itself. Overtime, emergency vendor help, and the engineering week you lose to the post-mortem.
- Team morale. Ask your ops team what they would pay never to see the outage channel light up at 2am again. Then look at your turnover numbers.
Run your own number
You don’t need a consultant. One formula, one napkin:
Hourly downtime cost = (revenue per hour × % of revenue that stops) + (staff affected × fully-loaded hourly wage) + recovery cost
A $6M-revenue business produces roughly $3,100 per hour. If an outage halts 80% of that and idles half the staff, you’re at $2,500–$3,000 per hour before counting recovery or reputation. Small numbers per hour — but check what a full day costs.
What actually shortens outages
Nobody credible promises zero downtime. What you can buy is a shorter, rarer outage:
- Monitoring with a human on call. You want to learn about the outage from your dashboard, not from an angry customer.
- Backups you have actually restored from. An untested backup is a hope, not a plan.
- Staged rollouts and feature flags. Most outages are self-inflicted; deploy small, deploy reversible.
- A one-page playbook. Who calls whom, where to look, when to roll back — decided in advance, at 10am, not 3am.
You can’t eliminate outages. You can make them rare, short and boring. That’s usually worth a lot more than it costs.