An uptime percentage is easy to compare and easy to misunderstand. Before treating two numbers as equivalent, ask what was measured, over what period, and what was excluded.
Turn the percentage into time
For a time-based availability calculation, the unavailable-time allowance is:
window duration × (1 − availability target)
Using a fixed 30-day window—not a calendar month of variable length—gives these illustrative budgets:
| Target | Unavailable time in 30 days |
|---|---|
| 99% | 7 hours 12 minutes |
| 99.9% | 43 minutes 12 seconds |
| 99.95% | 21 minutes 36 seconds |
| 99.99% | 4 minutes 19.2 seconds |
These calculations are not a ZnowPulse SLA or a promise about any monitored service. They simply translate percentages into time.
Successful checks are not the same as available seconds
If a dashboard divides successful checks by total checks, it is reporting a sampled success ratio. That may be useful, but it is not automatically an exact measurement of customer-visible downtime.
A short outage can occur between polls. Missing checks are another category: absence of an observation does not establish either health or failure. Read how your monitoring system handles them.
Decide what counts before the incident
Does planned maintenance count? What about one region failing, a slow but successful response, or one feature being unavailable while the homepage works?
There is no honest shortcut here. A number becomes useful when its definition matches the service customers depend on. Request-based availability may tell a different story from time-based availability, especially when traffic varies through the day.
Use the budget to make a decision
If the allowance for a critical service is only a few minutes, a five-minute polling interval cannot precisely resolve every interruption you care about. That does not mean the polling data has no value; it means the evidence has limits.
Read the check-interval math alongside your availability target. Then decide whether you need faster polling, application-level request metrics, or both. A more impressive percentage is less useful than a definition everyone understands.