In a previous blog post, I discussed how to determine high risk customers as it pertains to financial crime risk. To summarize, a risk professional can use publicly available information on crime statistics to augment the FFIEC guidance on establishing a BSA/AML Risk Assessment. Once high-risk customer segments have been determined, the next steps are gathering data and conducting analysis. There are varying methods of conducting analysis, but today I want to give an overview of a method called Peer Group Analysis.
Peer Group Analysis, in layperson’s terms, is the grouping of like businesses and entities for the purpose of comparison. From this grouping of like entities, various ratio and statistical methodologies can be applied to determine outliers. Once outliers have been isolated, analysis of these outliers is integral to learn more about the entity and their activity. Ultimately, the goal is to learn more about the business sector or determine whether the outlying activity is a consequence of suspicious activity. Once suspicious activity is determined, the associated data can provide basis for developing behavioral typologies to find other potential suspicious activity.
Now that Peer Group Analysis has been covered, I will walk through an example of a simple use case. Consider a credit union that maintains relationships with several pawn shops. Given the nature of pawn shops and the potential for high-risk activities such as dealing with precious metals and gems, short-term loan activity, and bulk cash handling, the credit union considers pawn shops to be higher risk. Once the pawn shops have been identified, the risk professional can then conduct a sampling of transactions using the same time frames, and then parse the data by transaction type.
Once parsed, the data can then be aggregated into various transaction type pools. The first thing to note is, while pooling the transactions, if the risk professional notices a transaction type that is not common to the peer group, isolate those transactions and determine causation. Next, various methods can be used to determine outliers, but for this instance, I will begin by using the Sample Standard Deviation formula found in Microsoft Excel as function STDEV.S.
Once I have calculated the sample standard deviation, I can apply it to my data and analyze it using the Empirical Rule. This rule, while not perfect, makes the statement that within a normal distribution of values, roughly 68% of values lie within one standard deviation of the mean, about 95% within two standard deviations, and about 99.7% within three. Knowing this, the sample standard deviation can then be applied against the pooled transaction data at either two or three standard deviations from the mean or average of the data sample. I can then assume that anything outside of my choice can be considered an outlier for further review.
While not perfect, this is one method of looking at high risk relationships and determining risk within their transactional data. Harnessing the ability of business tools such as Microsoft Excel makes using these powerful statistical methods a much easier task. If you are curious to learn more, feel free to reach out for a discussion!