GUEST BLOGGER
Erik Halvorson
Special Agent, U.S. Department of Energy

After working as a Special Agent for the U.S Air Force’s Office of Special Investigations for a couple of years, I was assigned to an overseas billet on the International Contract Corruption Task Force. The task force’s goal was to identify and stop corrupt U.S. government employees who were involved in the administration of federal awards taking place in the Middle East. My office was located in Riyadh, Kingdom of Saudi Arabia, where I was to be the sole primary agent in the country.

However, there was one problem: early in my career I was, in every sense of the word, a blunt instrument. I was completely and utterly uninitiated in one of the most important facets of investigations — technical proficiency. If there was a problem, I would try to smash through it. There wasn’t a need to think several moves ahead because I had a badge and a will to get the job done. Consequences be… well, you get the point.

Upon landing in the country, I received my first allegation. It was related to a construction project where the government’s technical representative, an American federal employee, was accepting bribes. The technical representative would write in contract modifications and up-spec the project so the company could make more money. He would then approve the work even though it never happened and take a cut of the profits. According to a cooperating witness, this government employee was then paid in cash, which he would deposit at a U.S. installation finance center and wire back to his American bank account.

Sounds simple enough. I figured the easiest way to approach this problem was to get copies of the paperwork for the cash deposits and copies of the company’s invoices to the government. Piece of cake. What I was not prepared for was the government’s records being even more disorganized than my own. I was led to an abandoned building in the desert that was filled from floor to ceiling with partially labeled boxes. They appeared to have either been thrown in by a very disgruntled employee or blown in by some form of natural disaster.

I began to go through both the financial documents and the contract documents. I had no knowledge of any technical approaches except to manually schedule everything out by hand and later transcribe it to Excel. I remember that tedious process to this day, now nine years later. I was sitting in that building without air-conditioning, in the hottest weather I have ever experienced. All I could think was, “There has to be a better way.”

I went home that night and found Benford’s Law, the law of anomalous numbers. Later that week, I found the ACFE and thus began my decade-long spiral down the rabbit hole of fraud analytics.

Since then, I’ve completed a master’s degree in applied analytics and began a Ph.D. in research methods and statistics. I’ve helped create proactive reviews that use analytics to successfully identify fraud and spin off federal investigations related to my department’s federal awards. I worked to help create analyses that systematically review large swaths of data that were previously ignored. Presently, I am also assigned as part of a nationwide team that uses analytics to identify and support investigations with some of the smartest federal employees I have ever met. During these past nine years of learning, I have come to rely on three steps that consistently pay off. The three steps are idea mapping, visualizations and statistical analyses. They are how I generally organize my initial review when I am beginning a potential project. Here’s how you can use them in your own cases.

Step 1: Idea mapping

The goal behind the first step, idea mapping, is to creatively evaluate your problem. Real solutions are found when a person is able to objectively review and define the problem, look at their resources and attack the problem creatively. Fraud analytics is no different. When I begin to design a proactive review, I look at:

the program requirements
the general schemes I want to target
the red flags those schemes produce

Then I will “idea map” data sources or records that will help me identify those red flags at scale. This is an iterative process. And not every scheme or project is conducive to an analytics analysis. Some judicious initial planning will save a lot of heartache on the back end.

Step 2: Visualizations

The second step, visualizing your data, ensures that your data is much easier to evaluate at scale. This is true for mapped or geographic locations, pattern analysis, identifying outliers, trends over time or distribution/scatter plots. As human beings we are predisposed to absorb and process complex data much easier when it is visualized. Our brains naturally perceive approximately 83% of stimulus through sight. This should be used to our advantage, not just in analysis but in briefings, interviews and even when testifying.

Step 3: Statistical analyses

Finally, the third step is to conduct a basic statistical analysis. Until you have done this on your data, it is borderline impossible to really, deeply understand it. A statistical baseline establishes the actual quantified trends and allows you to move past assumptions. This matters because often our initial assumptions are wrong or unsupported. Also, this independent analysis will immerse you in your data in a way that permits you to walk amongst it and understand it deeply. You can identify items of interest within the data, which is imperative for supporting operational aspects like interviews or legal briefings later on.

I would love to say that the case outlined above was successful, that the bad guy went to jail and the hero (me) was covered in laurels. However, my story is unfortunately a cautionary tale. There were several reasons why my case was not prosecuted — some were in my control and some, unfortunately, were not. However, I can’t help but think if I had approached it from a technically proficient standpoint and used some advanced analysis techniques it would have ended differently. Idea mapping and visualization techniques would have allowed me to present the story in a cleaner, smarter way, and a deeper analysis would have bulwarked my case work.

I was once told by a mentor of mine, and the greatest investigator I have ever known, an investigator’s only two modes cannot be: I forgot my tools at home or I am going to hit them over the head with my toolbox. In that spirit, fraud analytics is not the only tool that investigators need, but it is one of the most important and diverse skill sets we can develop.

This importance will only increase as the velocity, volume and variety of data increases within our work. Researchers hypothesize that by this year, each person on earth will create roughly 1.7 megabytes of data every second. In fact, way back in 2013, scientists hypothesized that more than 90% of humankind’s data, going back to the dawn of mankind, was created in the previous two years. This data will force us to continue to refine and learn new skills to stay ahead of fraud. Yet, this environment creates a unique opportunity for us to blaze new trails and define the future of analytics within our profession.

Featured