Skip to content

Blog

An error budget for small teams: spending your 43 minutes a month

An uptime target is easier to live with as a budget. At 99.9% a 30-day month allows 43 minutes 12 seconds of downtime. The question is what you spend it on.

Updated 25 October 2026 · By the Upcertain team

From target to budget

TargetBudget per 30-day monthPer week
99.5%3 h 36 m50 m 24 s
99.9%43 m 12 s10 m 5 s
99.95%21 m 36 s5 m 2 s
99.99%4 m 19 s1 m 0.5 s

The uptime calculator works out any other target and period.

What spends the budget

  • Deploys that fail briefly. Twenty deploys a month that each fail for 60 seconds spend 20 minutes, almost half of a 99.9% budget.
  • Incidents. One 30-minute outage spends most of the rest.
  • Slow detection. The minutes between failure and someone noticing count too. A 5-minute check interval can cost several minutes per incident before anyone starts work.
  • Maintenance, unless your target excludes planned windows.

Measure it

You need two things: monitoring that sees short failures, and incidents that start at the first failure rather than the moment someone noticed. Check every 30 seconds to 1 minute so 1- and 2-minute outages are counted at all; see check intervals. Upcertain records each incident from the first failed check and reports uptime over 24 hours, 7, 30 and 90 days.

Confirmation checks keep short failures from alerting anyone, and failures that never become incidents do not open an incident record. If you want deploy blips in the budget, look at the monitor's check log as well as its incidents.

Spend it on purpose

  1. At the start of the month, the whole budget is available for shipping.
  2. When an incident spends more than half of it, slow down risky changes until the end of the month.
  3. When the budget is gone, ship only fixes that make things more reliable.
  4. When a month ends with most of the budget unspent, you can afford to move faster, or promise customers a higher target.

Make deploys cheaper

For most small teams the cheapest minutes to win back are deploy minutes: rolling restarts, readiness checks and connection draining. The deploy guide lists them.

Publish what you promise

Whatever target you choose, a status page with real uptime history keeps you honest with customers and with yourselves. It is harder to argue that last month was fine when the bars say otherwise.

Questions

Does maintenance count against the budget?
That depends on what you promise. Many SLAs exclude announced maintenance. In Upcertain, time inside a maintenance window counts as maintenance, not downtime, so it does not lower the uptime figures.
What if one service has a different target?
Give each service its own target and budget. A marketing site at 99.5% and a payment API at 99.95% spend very differently.
Should we share the budget with customers?
Share the target and the real uptime history on your status page. The budget itself is an internal tool for deciding when to slow down.