A system and method for tracking subject behavior and making object recommendations include a data collection module that captures subject behavior and provides behavioral context for the recommendations; a profiling module that extracts characteristics of subjects and objects from the behavior data; and a recommendation module, which uses the profiles and the behavior context to generate personalized content and which includes a behavioral recommendation module, correlating profiles or uses other unconstrained methods for matching objects to subjects; a business rule module filtering and modifying recommendations by applying application-specific business logic to defined attributes of the objects; and a promotion module modifying the scores from the recommendation module to bias the recommendations towards certain objects based on additional business goals. The system continuously samples and assesses the performance of a variety of candidate recommendation strategies and optimizes the selection of rules and profiling methods.