What is dynamic baselining and how does it work?

Dynamic baselining is the statistical process AppDynamics uses to continuously learn what "normal" looks like for a metric — most often a Business Transaction's response time — instead of relying on a static threshold someone has to guess and maintain by hand.

It works by collecting historical values for a metric within matching time windows (the same hour of the same day of the week, by default) and computing statistics like the mean and standard deviation for that window. A default health rule then flags a violation when the current value deviates from that baseline by a configured number of standard deviations, rather than crossing one fixed number regardless of time or traffic.

Because the baseline recalculates on a rolling basis as new data comes in, it naturally adapts to gradual shifts — a service that's organically gotten a bit slower as data volume grows over months will have its baseline drift along with it, which is useful for tracking gradual degradation but also means a slow, creeping regression can sometimes go unflagged if teams rely on baseline deviation alone without also watching absolute response-time trends.