Any person or thing that generates or participates in an event. Actors on digital services may be people who access your service directly, and are often thought of as "users" or "people." However, you might have more than one kind of "person" actor, such as when a rider actor hails a ride with a driver actor in a ridesharing service. Actors can also be real or virtual things – like devices, topics, accounts or “bots” that can also be thought to “behave” and about whom you may have the same kinds of behavioral questions that you have about people, such as segmenting topics on a social site by metrics on the frequency of comment war sessions. For these reasons, Scuba does not equate "behavior" with "user behavior."
Actor columns are identified at ingest time by explicitly making them shard keys. When an event source has multiple actor populations, such as the rider/driver scenario above, we implement them through multiple shard keys in the same event table (dataset).
The series of events associated with a given actor is considered their journey. The patterns found in and across actors' journeys are what we refer to as behavior, such as progress through funnels. Behavioral knowledge objects like funnels, sessions, cohorts, and actor metrics depend on actors being ingested as shard keys, which is why it is important to identify the actors required to support your critical analyses before ingesting data.