Identifying Fraud Using Abnormal Duplications within Subsets
The tests in this chapter are based on the assumption that excessive duplications within subsets are indicators of fraud and errors. Because there is always going to be some amount of normal duplication, and some level of abnormal duplication we have to review our results carefully to find the duplications that are there because of errors or fraud. Another way to focus on important results is to only look at duplications above a dollar threshold. The goal is to run tests where abnormal duplications are reasonably reliable indicators of fraud or errors.
The first test described in the chapter is a straightforward test to find duplicate records. Although duplicate payments with all fields being the same are rare in accounts payable, it is possible for duplicates to arise in other situations. For example, an employee might duplicate a purchase using a purchasing card. The second test is a little more complex in that we are looking for partial duplicates. The most valuable results from the partial duplicates test has been from identifying cases of the (a) same dollar amounts, (b) same date, (c) same invoice number, and (d) different vendors. These errors occur when the wrong vendor is paid first and the correct vendor is paid later. The third test quantifies the level of duplication within a subset. Subsets are then ranked according to their duplication measures. A formula is used to calculate the duplication ...