Z-score deviations, velocity spikes, and cross-region cluster events across all monitored regions
An anomaly on this page is a move that sits meaningfully outside a metric's own recent behavior - typically more than two standard deviations from its trailing mean. We score the size of the deviation (the z-score), whether it's part of a cluster (several correlated metrics all moving at once), and how fast it built up (velocity). The point is not to flag every drop - it's to filter out the noise and show you moves that are genuinely out of pattern.
How to read it. Individual anomalies are interesting but often benign. Watch for clusters - several anomalies in the same region or across correlated metrics appearing at the same time. That's what usually precedes a regime shift. Severity tiers (moderate / high / extreme) map directly to z-score ranges and we show the exact number so you can disagree with our thresholds.
Limitations. Z-scores assume a relatively stable base distribution. After a real regime change the window takes time to recalibrate, which can either hide genuine stress (new normal is the old anomaly) or flag everything as anomalous for a few days. Opt in to browser notifications for extreme anomalies if you want to be paged when the dashboard sees something it considers out of pattern.