A check interval is not a notification guarantee. It is the spacing between opportunities to observe a failure. That distinction helps you choose monitoring that fits the service instead of buying a number.
The simplest case
Assume checks happen exactly every T seconds, the outage continues until a check sees it, and the failure begins at a uniformly random point between checks.
The wait for the next check is between zero and T seconds. Its average is T divided by two.
| Interval | Average wait for the next check | Longest wait under this model |
|---|---|---|
| 5 minutes | 2 minutes 30 seconds | Up to 5 minutes |
| 1 minute | 30 seconds | Up to 1 minute |
| 10 seconds | 5 seconds | Up to 10 seconds |
These are arithmetic examples, not measured ZnowPulse performance or service guarantees. Scheduling, timeouts, confirmation rules, and delivery add time.
Short outages can fall between checks
Suppose an outage lasts 30 seconds and checks are five minutes apart. In the same idealized model, with a random start time and effectively instantaneous observations, the chance of a check landing inside it is 30 divided by 300: 10%.
Real checks take time and do not follow perfectly uniform schedules. The point is that an all-green polling history does not prove that every second was healthy.
Confirmation changes the timeline
Requiring two consecutive failed checks helps avoid reacting to a single transient failure, but usually adds another interval before confirmation. The trade-off is worth making deliberately.
A faster interval also cannot fix an alert that goes to an unattended mailbox. Think through the whole path: observation, confirmation, delivery, acknowledgment, response.
Match speed to impact
For a low-traffic information site, a longer interval may be reasonable. For a customer-facing API where minutes of interruption matter, faster observation can be useful. Keep request cost in mind when choosing the endpoint.
ZnowPulse offers five-minute checks on Free and intervals as short as ten seconds on Pro. Compare the current plans, then choose based on the failure you need to notice—not the fastest number alone.