The present invention relates to a subscription modeling system and methodology. A query processor receives subscription query and subscriber information and transforms the query and subscriber information into data. An index component propagates at least one database with the transformed data; and a matching component that associates the transformed data with event data to generate a database of notification data that can be delivered to subscribers. The invention provides for abstracting subscription and subscriber information (as well as event information if desired) to high-level classes (e.g., data fields)--thus the invention provides for modeling such notification related information as data. Subscription applications can thus be developed at high levels wherein complex subscription queries and subscriber information can be defined as data fields for example. Databases in accordance with the data fields can be propagated with subscription/subscriber specific information. The present invention takes advantages of the processing power associated with database engines (e.g., SQL server) to generate notifications via performing a join operation on the databases (e.g., subscription database(s), subscriber database(s) and event database(s)). Accordingly, notifications are generated en masse as compared to per subscription per subscriber. Thus the present invention provides for a highly scalable and efficient notification system.