In this paper we propose a data analysis based audit framework where we identify and eliminate clusters of data points that do not have the characteristics of a Benford conforming data set. By looking at the attributes of these data sets, we identify potential audit candidates iteratively with the objective of utilizing the auditing budget in a more efficient way. We analyze a publicly available real data set that contains a list of contracts belong to a public health care organization using the proposed framework. We believe this systemic approach is better than a random selection process to better utilize audit resources.