Alembic uses a statistical technique called anomaly detection to identify variations in time-series data for a given metric. It takes all your activity for a metric–say, Facebook impressions–and computes a median for that activity over a rolling time window. When activity within that metric spikes to above a level that is noticeably different from the median, this is called an anomaly. Alembic refers to these spikes, or anomalies, as detections.
What does Alembic mean by "detection"?
