How organizations are using analytics to prevent fraud

People commit fraud in a number of different ways:

  • A cybercriminal could steal your debit card number.
  • A member of an organized crime group could launder illegitimate money through a legitimate business. 
  • An investor can orchestrate a Ponzi scheme.
  • A person posing as a telemarketer could get you to divulge your bank account information.

It's a scary prospect to think about, but organizations of every ilk need to take seriously. The question is, how are enterprises identifying fraud when it's committed? Better yet, how are they using technology to help them prevent it from happening in the first place? 

"While the type of fraud committed can depend on the insurance plan a claim holder has purchased, insurers often use legacy systems in a troublesome attempt to scrutinize a person of interest's relationships."

The insurer's dilemma 
Insurance companies are well aware of how tempting it is for some of their customers to submit false claims. While the type of fraud committed can depend on the insurance plan a claim holder has purchased (health care, property, etc.), insurers often use legacy systems in a troublesome attempt to scrutinize a person of interest's relationships. 

Information Age's Ben Rossi noted that data from third-parties and suppliers are critical components that can't be ignored, either. Data aggregated from outside sources such as social media, other insurance firms and employers can tell a lot about a person's behavior, habits and past discrepancies. However, gaining access to these repositories isn't always easy as far as protocol is concerned. 

Yet the greatest factor preventing insurers from gaining accurate insights into how and why a person is committing fraud (or likely to) is the prevalence of legacy solutions. Rossi maintained that programs designed to create red flags exacerbate inefficiencies. The reason why these systems are such a bother is because the alerts they produce often provide little intelligence regarding a transgression. 

The insurer's definition of "advanced analytics"
For a professional working in the insurance industry, "advanced analytics" describes a solution that is capable of outlining specific behavioral trends that strongly suggest an individual is likely to perpetrate a scam. Rossi explained how these qualitative data analysis programs work:

  1. The software collects information from every department within the insurance firm, refining it if necessary. 
  2. The solution employs several business rules that provide it with a way to discern whether fraud is persisting.
  3. From there, it draws connections between various data points, emphasizing patterns or trends that suggest vulnerabilities. 
  4. When certain persons of interest are identified, the company can send out information requests to third parties.
  5. Once they receive approval and/or reports from outside entities, the company enters the data into the program, which employs text and sentiment analysis to ascertain details of an individual's personality. 

While this process not only gives insurers a comprehensive understanding of their customers, it also enables them to identify processes and protocols that are allowing fraud to be committed in the first place. 

"There's one algorithm that can help financial experts identify fraud, and it's name is Benford's Law."

What's the math behind it? 
Sure, data analytics can perform some pretty amazing feats. Yet behind every data analysis program lies a set of instructions, or algorithm, designed to process information. In regard to accounting, there's one algorithm that can help financial experts identify fraud, and its name is Benford's Law. 

The Wall Street Journal's Jo McGinty explained Benford's Law. Basically, the rule states that 30.1 percent of numbers in a collection of financial transactions begin with "one." Each proceeding digit (two, three, four, etc.) should "represent a progressively smaller portion."

A data visualization developed by Dan Amiram of Columbia University displays the anticipated Benford frequencies, which forms a long curving slope. A dotted line marks each number's value. If the dots stray away from the orange line (either higher or lower), it's an indication that fraud has probably been committed. 

Although Benford Law serves as a basic measurement, it must be noted that deviances within a Benford graph should not serve as the definitive conclusion that a business or person has perpetrated fraud. However, it does provide a good example as to what kind of formulas data analytics programs are using.