Automated metric monitoring
AI Anomaly Agent
Point it at an Excel or CSV sheet of business metrics. It builds a baseline, flags unusual movement, explains the change in business language, and emails an alert when something is off.
"Normal" is learned from the last 7 rows before the newest one. The agent averages those 7 values to get a baseline. A short window (3–5) reacts fast but is jumpy; a long window (14–21) is stable but slow to notice a new trend.
A z-score says how many standard deviations the newest value sits from the baseline. At 2.0σ a metric is flagged only when it moves further than usual noise. Lower it to catch more (more false alarms), raise it to only see big shocks.
Every metric starts on the global settings. Use the panel below to give a noisy metric like Traffic a longer window or a stricter σ, tighten a critical one like Revenue, or switch a metric off so it never triggers an alert.
Drop an .xlsx, .xls or .csv file here
First sheet, one date column plus numeric metric columns.
Global baseline window
7 periodsGlobal sensitivity (z-score)
2.0σBaseline = average of the previous 7 periods. A metric is flagged when the newest value sits more than 2.0 standard deviations away from that baseline. Override any single metric below.
Recent alerts
View full historyNo alerts sent yet.
