What kinds of employee fraud does AI catch?
Employee theft and fraud in the retail industry has been well-documented for decades. The most recent National Retail Security Survey (NRSS) from the National Retail Federation places the 2018 cost employee fraud at $1,203.16 on average with a median of $700. Technology solutions make it easier for retail loss prevention and asset protection departments to find, substantiate, record, and resolve employee fraud across the retail organization.
Exception based reporting (EBR) solutions use advanced data analytics to examine the retailer’s POS data to find unusual behavior. A little more than half of the retailers use exception based reporting (EBR) to examine their POS data according to the NRSS report. This technology revolutionized exception reporting in the early 2000s, but modern options are now available.
Artificial intelligence (AI) enables the EBR solution to make basic decisions without human intervention. Decisions like, “Is this something I should look deeper into”, “Is this case worth pursuing,” or “What steps should be taken next?” AI helps to avoid false-positives that sidetrack retail investigators and its problem-solving technology helps end users discover and resolve more cases in less time, improving departmental performance and protecting company profits.
At present, the types of employee fraud AI discovers are usually tied in some way to a retail transaction, whether that is at the point-of-sale or in the back room. These can include return fraud, sweethearting, collusion with other employees or consumers, loyalty program abuse, merchandise theft, and cash theft.
The AI system connects the dots for the end user, correlating a variety of activities and searching for patterns of behavior. When a pattern is found that contains enough fraud-like indicators, the system summarizes the situation and flags the end user to investigate further.
How does AI compare?
Appriss Retail conducted an experiment with one retailer. Using 12 weeks of data that had already been screened and investigated by a conventional EBR system, Appriss Retail screened it with an EBR system enhanced with AI. Two rounds of testing were performed to evaluate the system’s ability to identify previously unidentified cases and to determine the quality of the cases being provided. Through two rounds of testing, there was at least a 25% increase in fraud-related terminations above what would normally be captured by EBR alone, and additionally, a 25% decrease in time required to track down the fraudulent activity within many of the cases. The system found 14 more cases, most of which resulted in employee termination. Among the frauds found were:
- Non-receipted returns rung to the employee’s credit card
- Items purchased at a discount refunded at full price
- Giving merchandise to friends and family
- Using receipts that were left behind by consumers to create refunds which the employee kept
Over time, a system using a feedback loop becomes more intelligent and attuned to the types of employee fraud that the retailer acts on. Action is an important feature. The goal of AI is not to inundate investigators with an endless stream of infractions. Rather, it provides investigators with the cases that they are likely to resolve in some fashion.
According to the National Retail Federation, 72 percent of employee fraud cases are not pursued to prosecution or restitution. AI helps to improve performance by focusing resources on the employee fraud cases that are most likely to be productive.
“The $35 billion in loss due to retail crime is the single largest category of property crime each year,” states the Employee Fraud Special Report from LP Magazine. Employee fraud is a large and persistent portion of that loss. AI has proved to be a valuable tool in finding previously-undetected fraudulent behavior.