Three Fraud Factors and How to Find Them
This post is intended for those professionals who have been tasked with identifying some form of occupational fraud present in their respective industries. Not sure if you apply?
Luckily, there is a one question test to determine if your trade is on the list:
Does any part of your job involve time, people, and money?
If the answer is yes, you exist in the physical universe we all agree to be reality, and fraud is a part of it.
If you think the answer is no because of how one categorizes “money”, please consider that the question does not exclude individuals without an income, or who trade in abstract currencies such as the debt market or executive power grabs. We won’t waste words debating if you are human to refute the “people” portion of the query and go straight to an explanation as to why “time” is a fundamental factor of fraud analysis.
Occupational fraud is rarely a singular event. For those familiar with the common accounting practice of double book keeping, each transaction has at least two records of offsetting debit and credit amounts. Corruption, theft, and the financial statement schemes can all be reduced in principle to an intentional break in the normal sequence of transactions by an individual for personal gain.
Sequence = Actions or Measurements over Time.
Before your half-empty glass tips over in despair, please take a moment to remove any predisposed ideas you may have about the people who commit fraud once, a few times, or as their full-time profession. Fraud is not a question of good or evil, but rather an inevitable factor of human social behavior, which includes some level of discernable routine and seasonality.
In this article, I will reveal Three Fraud Factors that you apply to any data set to better understand and visualize the potential impact of fraud on your business.
Time in the world of fraud data analytics is focused on the regular intersection of real time, and transaction sequences. Order numbers register transaction or receipt numbers, P&L statement draft version numbers, employee ID’s, check numbers, and even the ticket you grab at the grocery store deli counter are examples of transaction sequences that can be analyzed against real time to identify fraud. Once you have these values, visualize seasonal trends into continuous and discrete time groups at all available splits of time from years to single minutes if possible. The expression of where these sequences and time intersect is the key to finding fraud. Create a few dashboards with the sequence vizzes as a reference for your future analysis.
Example 1. Sequence viz
2. Social, not Individual Behavior: create groups instead of looking for lone actors
a. Create a 4-axis scatter plot with 4 different KPI’s and set all the axis’ so that the outliers migrate towards the middle of the view. Then, add in average or median lines for all the axis’ shown. You have now created a target zone for visualizing and analyzing outliers.
b. The next step is leverage Tableau’s clustering features and create at least 5 groups. Resist compartmentalizing by organizational structure (districts, regions, report-to chains) for now, and just focus on creating clusters of points at varying levels of detail.
c. Finally, create a set by selecting all the marks inside of the average box you input earlier. Feel free to rename the two resulting sets In and Out or Above Avg. and Below Avg. Don’t worry about investigating potential cases yet; this process is just about slicing the data into more manageable groups. If a few bad apples can spoil the whole barrel, we are simply putting the same volume of apples into and increasing number of barrels with a relatively decreasing capacity until we start to notice some barrels that are rotten than are market ready. There may be occasional cases where a single perpetrator devised, enacted, and solely benefited from a scheme, but they are extremely rare.
3. Money: Plan to Steal Something, then Define the Sequence in your Data
a. Begin with a dollar amount and a deadline, then get creative with the circumstances. For example, you need $10,000 more than you have in all your combined assets to pay for a heart transplant that must be done in the next 60 days. Fraud requires some level of motivation to overcome the potential consequences if the perpetrator is caught.
b. Fraud schemes can be extremely creative, but most prolonged methods of deception and concealment are a series of learned behaviors. Whether directly taught to a coconspirator or learned through indirect experience, there is a laundry list of internal and external scams for each industry. Research known fraud types in your profession and spend a significant about of time modeling a hypothetical fraud case. If possible, plan a real fraud scenario against your company, but do tell someone first so you don’t end up on your own case list. Document every single step, exactly how the actions could be concealed, and at what point you found a control you were not sure how to circumvent.
c. Lastly, define the data points and metrics that currently exist, or can be created, to gain visibility of the scheme you devised.
d. Repeat this exercise as often as you can while mixing up the urgency and target amount. $100,000 over 5 years, $25 each day, enough to pay a mortgage each month, etc. Be sure to invest a significant amount of time hypothesizing the personal circumstances to build a diverse list of testable scenarios, and as a deliberate method of humanizing your subjects.
The result is an active and continually adapting fraud analytics strategy that leverages data visualization to uncover intentional deceptions to the normal accounting of time, people, and money through your business.