Thresholds

checking…
U
user #0 · admin

Thresholds

Safe / fail bands per parameter. Auto rows are seeded from the profiler's percentiles; manual rows are operator overrides.

Asset Parameter safe_min safe_max fail_min fail_max Source Updated by Updated
loading…

Thresholds — how to read this page

What you're looking at

The "safe band" for each (asset, parameter). For a given reading we ask two questions:

The forecasting / anomaly engine uses these bands together with the model's own prediction to decide whether to raise an event.

How the bands are seeded

For every numeric series the profiler picks the 1st and 99th percentile from the last 30 days of history and writes them as the default safe_min / safe_max. That's the auto source.

An operator can override these in the UI (coming in Phase 5) or via API. Overrides are tagged manual and the auto-seeder will never overwrite a manual row.

Columns explained

How to use this page

Use it to check what numeric range the service considers normal for a parameter. If anomalies are firing too eagerly for one signal, the safe band is probably too tight — widen it via a manual override. If the band is much wider than reality, real failures could slip through unflagged.

A row missing fail_min/fail_max simply means no hard fail limits have been set; only the safe band is in use.