Remote Monitoring Systems: Sensors, Alerts and Dashboards

Remote monitoring is one of the most common IoT applications: temperatures in cold stores, levels in tanks, vibration on pumps, power in remote cabinets. The parts are familiar, a sensor, a connection, a database and a dashboard, but systems succeed or fail on details: whether readings can be trusted, whether silence is noticed, and whether alerts are rare enough that people still act on them.

At a glance

Heartbeat
A regular message proving the device is alive, so silence itself can raise an alert.
Hysteresis
Separate trigger and clear levels stop an alarm chattering around a threshold.
Delay
Requiring a condition to persist before alerting filters out brief spikes.
UTC
Storing timestamps in UTC avoids confusion across time zones and clock changes.

The problem

The common failure is not a missed reading but a missed absence. A device that loses power or signal stops sending, and a system that alerts only on bad values reports nothing, so a failed freezer and a failed sensor look the same: quiet.

The opposite failure is alert fatigue. Thresholds set without hysteresis or delay generate repeated notifications for marginal conditions, people learn to ignore them, and the one real alarm is ignored with the rest.

Treat silence as a reading. A system that cannot tell a quiet site from a dead sensor is not monitoring.

Designing alerts people act on

Each alert needs a clear condition, a persistence time, a separate clear condition, an owner and an escalation path if nobody acknowledges it. Missing data, low battery and sensor faults deserve alerts of their own, routed to whoever maintains the devices.

Reporting interval sets both responsiveness and running cost; the first figure shows messages per device per day at common intervals. Many products report on a slow interval and send immediately when a reading crosses a threshold, keeping traffic low without delaying alarms.

What good looks like

For each measured value, define the normal range, alarm conditions, persistence and clear levels, and who must be told.

Add heartbeat and data-quality monitoring for every device, with alerts that reach the people responsible for maintaining them.

Review alert history regularly and tune any alert that fires often without leading to action.

Messages per device per day
messages
Every 10 s8640
Every minute1440
Every 5 minutes288
Every 15 minutes96
Every hour24
Calculated. Multiply by fleet size and message size to estimate data and platform costs. Logarithmic scale.

From reading to response

  1. Measure: Reading with timestamp
  2. Check: Range, persistence, data quality
  3. Alert: Notify the owner
  4. Escalate: If not acknowledged
  5. Clear: When the condition resolves
Each stage needs a defined rule. Missing data enters at the check stage like any other condition.

What it changes

Trusted Alerts

Notifications are rare and meaningful, so people respond to them.

No Silent Failures

Dead devices and lost connections are detected as quickly as bad readings.

Controlled Costs

Reporting intervals match the response time the application needs.

Useful Dashboards

Views answer operational questions rather than plotting every value.

Where to go next

Calibrate sensors and record calibration dates. A well-designed alert on an inaccurate reading is still wrong.

Plan retention and downsampling: keep raw data for a defined period and longer-term summaries for trends.

Problems this prevents

  • Dead sensors mistaken for quiet sites
  • Alarms chattering around a threshold
  • Alerts with no owner or escalation
  • Timestamps confused by time zones

Discuss your project

Email support@aavico.com or use Get in Touch when the website finishes loading.