Health Metrics Explained
What Beacon's numbers actually measure, and how to read them without drawing the wrong conclusion.
How Often the Numbers Update
Network status checks run automatically every two hours, at half past the hour, and a daily summary is published at 08:00 EAT. Alert records are retained for 90 days.
Two consequences worth knowing:
- A monitor that failed twenty minutes ago may not show as offline yet. The dashboard is near-real-time, not live.
- History older than 90 days isn't available in the alert record, so long-range reporting should be exported as you go rather than reconstructed later.
Fleet Metrics
Online Rate
The share of monitors transmitting, derived from the proportion not transmitting. Only the fully offline Not transmitting state counts against it — see Device States.
Online rate is a snapshot of how many monitors are transmitting, not a measure of how reliably each one reports. For reliability, look at uptime — and remember that an average uptime hides distribution. Ten monitors at 90%, and nine at 99% with one dead, both average about 90%. The Uptime Range (Min–Max) column on the dashboard is what separates them, and it's usually the more actionable number.
Alert Levels
Fleet status is set by the percentage of monitors not transmitting:
| Status | Condition |
|---|---|
| OK | Under 35% not transmitting |
| WARNING | 35% to under 50% not transmitting |
| CRITICAL | 50% or more not transmitting |
Individual alerts carry a matching severity — LOW, MEDIUM, or HIGH — on the same thresholds.
These thresholds describe the fleet, not a single monitor. A fleet at OK can still contain completely dead monitors. Never read "OK" as "nothing needs attention" — it means the network as a whole is functioning, which is a much weaker claim.
Incidents
The count of warning and critical incidents raised in the last 14 days. Useful as a trend: a fleet with a good current online rate but a high incident count is unstable, flapping in and out of service, and that intermittency is often harder to diagnose than a clean failure.
Per-Device Metrics
Available on a monitor's Performance tab.
Daily Uptime
Reporting consistency day by day. Read the shape, not just the number: a steady 85% and an alternating pattern of 100% and 0% average the same and mean entirely different things.
Data Frequency
How often readings arrive. Falling frequency with uptime otherwise intact usually points at connectivity rather than power or sensors.
Sensor Health
Whether the sensor is reading sensibly, expressed through correlation and error margin.
- Correlation — how well the monitor tracks what it should. Co-located monitors should move together; when they stop agreeing, one is wrong.
- Error margin — the size of the discrepancy.
A monitor can score perfectly on uptime and badly here. It is online, reporting punctually, and producing numbers you shouldn't use.
Battery Voltage
Power status, and the leading indicator for solar-powered units. Voltage sagging before uptime drops is the signature of a power budget problem — often a panel that has become shaded or dirty rather than a failing battery.
Diagnosing by Combination
No single metric identifies a cause. The combination usually does:
| Pattern | Likely cause |
|---|---|
| Battery falls, then uptime falls | Power — panel, shading, or battery |
| Uptime steady, sensor health degrading | Sensor drift or failure |
| Data frequency falls, battery fine | Connectivity |
| Fails overnight, recovers by day | Solar power budget |
| Several monitors in one area fail together | Environmental or network, not hardware |
| Stale feed but still transmitting | Integration or timestamp, not the field |
Use Uptime Heatmaps to spot time-of-day patterns. A failure that recurs at the same hour is nearly always power or scheduled network behaviour, and a heatmap makes that visible in a way a timeseries doesn't.
A Note on Comparing Monitors
Two cautions when ranking hardware:
- Compare like with like. Low-cost sensors and reference monitors have different expected error margins. Judging the former by the latter's standard makes healthy units look broken.
- Account for deployment date. Uptime spans the whole selected window, so a recently installed monitor is penalised for time before it existed on site. Narrow the range to since deployment.
What's Next
- Device States — What each status means.
- Device Data Analysis — Chart these metrics.
- The Fleet Dashboard — Where the fleet figures appear.