How the cost of downtime is calculated
The fastest honest estimate is pure arithmetic: take your revenue for a period, divide by the number of minutes in that period, and you have revenue at risk per minute. Multiply by the length of an outage and you have the cost of that incident. If only part of your business depends on the system that went down, scale it by that share. That's exactly what the calculator above does — and it deliberately counts only directly lost revenue, so the number is a floor, not a ceiling.
The reason the figure lands harder than people expect is that minutes are small and revenue is large. At $300,000 a month you're losing roughly $6.85 every minute the system is dark — about $411 an hour, before a single customer churns or a single SLA credit is owed.
Uptime, SLAs, and your error budget
An uptime SLA is a promise about how little downtime you'll have. The gap between that promise and 100% is your error budget — the amount of downtime you're allowed to "spend" before you're in breach. Three nines (99.9%) sounds airtight until you do the math: it's about 43 minutes a month. Four nines (99.99%) leaves you just ~4 minutes. A single bad deploy can blow a month's entire budget in one afternoon.
Most preventable downtime isn't exotic infrastructure failure — it's change. A bad deploy, a risky migration, a config mistake, or an AI agent running a destructive command against production with nothing standing in the way. That whole category is self-inflicted, which means it's also preventable.
How to stop spending your budget on self-inflicted outages
If change is what breaks production, then the highest-leverage fix is to put a gate in front of change. A control plane sits between everything that wants to modify production — humans, scripts, and AI agents alike — and production itself: deploys, restarts, migrations, and recoveries pause for your approval, run with least-privilege access, and write to a tamper-evident audit trail. The minute you'd have lost never starts.
This calculator is a 2-minute snapshot. Infraveil is the continuous version.
Infraveil is a control plane you run on your own servers. It watches the runtime, gates every production-changing action behind your approval, and gives you one-click recovery with audit-grade proof — so the outages you just priced out mostly never happen. Read the runtime. Approve the fix. Prove what happened.
See the live demo →Frequently asked questions
How do you calculate the cost of downtime?
Cost per minute = revenue for the period ÷ minutes in that period, scaled by the share of revenue that depends on the system that's down. Multiply by the outage length for the cost of an incident. It's a revenue-only floor — real cost runs higher.
What's a good uptime SLA?
It depends on what an outage costs you — which is exactly what the tool shows. As a reference: 99.9% allows ~43 min/month, 99.99% allows ~4 min/month. Pick the target where the cost of more nines is worth less than the downtime it prevents.
Does this include churn and SLA penalties?
No — it's intentionally conservative and counts only directly lost revenue. Churn, SLA credits, support load, and engineering firefighting time stack on top.
How do I actually reduce downtime from deploys and AI agents?
Put a human-approval gate between anything that changes production and production itself, scope access to least privilege, and keep an audit trail with one-click rollback. That removes the single largest category of preventable, self-inflicted downtime. That's what Infraveil does →