Other data analysis techniques
On Steps 1.10 and 1.11, we explored Benford’s Law in detail and discussed how this can be used to detect fraud. You’ll have an opportunity to put this technique into practice when you complete the course assignment in Week 3.
However, as Professor Evans explained in the video, there are several other data analysis techniques which can be used to find potentially fraudulent data. Here are a few examples.
Outliers are figures in a data set which are much higher or much lower than the normal range of the data. These can occur naturally, but they may also be a sign of fraudulent activity. Outlier investigation allows a fraud examiner to find these abnormal figures quickly and reliably, and identify results which need to be investigated further.
The preview of this computer science study explains how the researchers used outlier investigation to detect health insurance fraud.
This technique compares sets of text-based data, such as employee and vendor addresses. Fuzzy matching will match sections of text which are similar, but not 100% identical. It can be an effective way to find fictitious vendors which have been created in an accounting system: for example, a fraudster may make payments to a dummy company called A&TT, hoping that his or her employee will not spot the fraud because the vendor’s name is so similar to that of the real telecoms company, AT&T.
This description of a software product provides some other examples of how fuzzy matching could be used.
Time trend analysis
Time trend analysis examines how the cost, quantity or price of something has changed over time. If one of these values has increased abnormally over a certain period, this could be a sign of fraud. Time trend analysis can be used to examine large sets of data, and identify potentially fraudulent trends which require further investigation.
© Kogod School of Business, American University