The short answer
For every alert, answer three questions first: who receives it, what do they do, and does it justify waking someone. Alerts with no answer to the third should not fire outside hours.
Monitoring systems fail through volume rather than through gaps. The team stops reading the channel, and then the one that mattered is missed.
What continuous monitoring covers
- Plant and equipment status from the building management system
- Temperature and humidity in areas where it matters
- Water leak detection, which is the classic high-value case
- Door and access events out of hours
- Energy consumption anomalies
- Sensor and device health, so silence is distinguishable from normality
Leak detection is usually the clearest business case. The cost of a leak found in the morning against one found at two in the morning is not comparable.
Tier the alerts
| Tier | Example | Response |
|---|---|---|
| Wake someone | Water detected, fire system fault | Call out, immediately |
| Morning priority | Plant running out of range overnight | First task of the day |
| Log and review | Minor deviations, single readings | Weekly review |
| Trend only | Gradual efficiency drift | Monthly, no alert |
Most alerts belong in the bottom two rows. Getting that split right is what keeps the top row credible.
Persistence beats instant
A single out-of-range reading is usually noise: a door held open, a sensor glitch, a cleaning trolley in the wrong place. Requiring a condition to persist for a period removes most false alarms at the cost of a short delay.
Keep a small number of genuinely instant alerts for conditions where minutes matter, and make everything else require persistence.
Name the responder
- Every out-of-hours alert has a named person or a contracted service, not a shared mailbox.
- The alert says what to do, not just what happened.
- There is an escalation path if the first person does not respond.
- Every call-out is recorded, including whether it was justified.
- Alerts that never produce useful action get removed at review.
The last point is what keeps the system alive over years. Alert sets grow and nobody prunes them unless it is scheduled.