This document proposes a new approach to enable users to directly search vast datasets for potential items of interest. Current search methods for databases, data mining, and information retrieval have limitations when applied individually to today's large datasets. The proposed solution adapts elements from these different methods by using relevance estimation concepts from information retrieval, a multidimensional utility function from utility theory, and query refinement through explicit and implicit user feedback from data mining. This novel approach aims to decrease the difficulty of finding important data by combining human and machine intelligence to better support ad hoc searching of multidimensional data.