Sigma-based anomaly detection
Last updated 2026-06-14
Definition
Sigma-based anomaly detection flags a metric reading as unusual by measuring how far it sits from its own recent baseline, counted in standard deviations. Quri learns what a normal range looks like for each metric, then alerts you when a value drifts past that band, so a real spike or dip stands out from ordinary noise.
How to do this in Quri
- Connect a source and let Quri build a baseline from your metric history.
- Leave detection running — it watches in the background and surfaces flags on your Pulse home.
- Open an alert to see the value, its baseline, and how many sigmas off it landed.
- Try the idea first without an account at /tools/anomaly-spotter or /tools/metric-health-check.
Frequently asked
- What does “sigma” mean here?
- Sigma is the standard deviation of a metric around its baseline. A reading two or three sigmas out is far enough from normal that it is worth a look, which is what Quri flags.
- Do I have to set the thresholds myself?
- No. Quri builds the baseline from your own data and flags values that sit too far from it. You can try the same approach on a single metric at /tools/anomaly-spotter.
- Will normal ups and downs trigger alerts?
- Ordinary swings stay inside the expected band, so they do not flag. Detection fires when a value moves far enough past the baseline to stand apart from everyday noise.