Anomalies are detected when the actual cost of a resource suddenly exceeds the cost prediction past a predetermined trust margin.
- The app looks at the actual cost of all your resources every 24 hours and makes a cost prediction.
- The difference between the actual cost and cost prediction is calculated. This is called the prediction error.
- The app looks at your resource's prediction error trends for the last 30 days to establish a trust margin.
- If the actual cost suddenly exceeds your trust margin, an anomaly is flagged.
Note: The trust margin is raised for a few days following an anomaly detection to avoid multiple alerts on the same anomaly.