Logically, appropriately structured CER rules can be thought of as being made up of rules classes for two distinct logical functions:
Data rule classes are those which are an exact mirror for the data held by the application. For example Income Evidence would be a data rules class.
All other rule classes can be considered as Derived Data/Business Logic rule classes. For example Monthly Income would be such a rule class.
Additionally each rule class, provided out of the box, can be thought of as containing either:
Fixed processing which is relied upon by other processing in the application, and which cannot be altered; or
Processing which can be cloned and used as the starting point for your rule set work.