Sign in →

Metering Health

Per-metric metering health — event volume, error rate, latency, and Z-score anomaly detection so you catch a broken meter before it breaks billing.

Updated 2026-07-29Suggest edits

Metering Health

Metering Health (Operations → Metering Health) tracks the health of each billable unit's metering, so a meter that suddenly stops, spikes, or starts erroring gets caught before it corrupts billing.

Health states

StatusTrigger
HEALTHYNormal operation
WARNINGAn active anomaly, or error rate above 5%
NO_DATANo events received

Each metric card shows one of these three states. (CRITICAL is an anomaly severity, not a card status — see Anomaly detection below.)

The health grid uses fixed 24-hour and last-hour windows with a manual Refresh. Clicking a card opens a trend side-panel with a 7d / 30d / 90d period selector.

Per-metric signals

Each metric reports event count (24h), events in the last hour, error rate, average latency, P95/P99 latency, last-event time, and a recent-counts sparkline.

Anomaly detection

When detected, anomalies appear in an alert strip and table. Each anomaly has a type (SPIKE / DROP), severity (CRITICAL / HIGH / WARNING), sigma deviation, and observed vs expected values.

ℹ

Metering Health focuses on a single metric's pipeline health. For pipeline-wide throughput and source health, see Ingestion.