Event sourcing is an approach that tackles reproducibility as a shortcoming of CRUD-based systems.

Event sourced systems calculate the current state of the systems from atomic events that happened in the past. The events represent the individual business use case invocations, including the information provided in the invocations.

The current state is not permanently persisted, but emerges by applying all events one after another. The events themselves happened in the past and are immutable.

To give an example, a user with its characteristics is calculated from all events related to it. Applying UserCreated, UserApproved, and UserNameChanged one after another creates the current representation of the user up to its recent event:

The events contain self-sufficient information mostly concerning the corresponding use case. For example, a UserNameChanged event contains the time stamp and the name the user was changed to, not other, unrelated information about the user. The event's information is therefore atomic.

Events are never changed nor deleted. If a domain entity is removed from the application, there will be a corresponding deletion event such as UserDeleted. The current state of the system then won't contain this user anymore after applying all events.