01-28-2013 07:47 AM
Analytics is based on a simple principle: track and alert any abnormal behavior in the system.
Analytics identifies abnormal behavior by applying a dynamic baseline. The dynamic baseline is applied to each metric in the system; for example, each transaction from each location will get a different baseline which maps behavior of that specific metric. The baseline is calculated every week but takes into account activity over the last month.
There is a unique aspect to Analytics’ baseline calculation: seasonality. The algorithm identifies if a metric has seasonal behavior, for example does it behave differently during working hours and off line hours? Are there typical weekend patterns? Does it have periodic behavior that repeats every 3 hours? If Analytics identifies seasonal behavior, it will apply a different threshold based on the time and day.
As a different baseline is calculated for each metric, the end result is a dynamic baseline that can determine expected behavior for each metric at any given time.