Retail fraud and loss prevention have always been significant business concerns impacting profitability and customer trust. However, with the emergence of AI technologies, there is a newfound potential to combat these challenges more effectively.
AI offers sophisticated tools and algorithms to analyze vast amounts of data in real-time, enabling retailers to detect fraudulent activities and prevent losses.
Another tool helping retailers in asset protection is standardizing reporting and information. In an upcoming ‘AI in Business’ podcast interview with Emerj, Adam Oberdick, Lead Director of Asset Protection at CVS Health, tells Senior Editor Matthew DeMello:
“Ten years ago, you would absolutely say that every retailer is capturing information a little bit differently. And that makes it really difficult for law enforcement and prosecuting attorneys to put a collection together to actually go after a group,” says Oberdick.
“And so now you’re starting to see the standardization of resources as well. So I think that those things have really changed in the way that we’re looking at information in the way that we’re capturing those things,” he continues.
Emerj recently spoke with Chris Nelson, Senior Vice President and Head of Asset Protection at Gap Inc., on Emerj’s ‘AI in Business’ podcast to discuss the importance of data analysis and pattern identification for fraud prevention and the potential of artificial intelligence in enhancing security systems.
In the following analysis of their conversation, we examine two key insights from their discussion:
- Categorizing organized crime’s impact on profitability: Using anomaly detection for measuring the impact and behavior of worst-case offenders of fraud that are often the leading cause of losses in the retail sector, and without hurting sales.
- Harnessing AI and data for fraud detection and prevention: Differentiating fraudulent behavior through data analysis and pattern identification is crucial to identifying and addressing worst-case offenders while balancing the impact on overall operations.
Listen to the full episode below:
Expertise: Private sector loss prevention, law enforcement
Brief Recognition: Chris started his career as a commissioned officer in the US Army’s Military Police Corp. He transitioned his career to the private sector, joining Target Corporation’s Assets Protection team. Before his current role, Chris served Gap Inc. as the Senior Director of LP for Banana Republic, as Vice President of Loss Prevention for the Old Navy Brand, and most recently as the Head of Loss Prevention (interim), leading all of Gap Inc.’s loss prevention, security, and business continuation program (BCP) teams and efforts before being promoted to Senior Vice President and Head of Asset Protection.
Categorizing Organized Crime’s Impact on Profitability
Chris begins by highlighting the widespread nature of organized crime. He notes that significant company CEOs now acknowledge loss as a material impact on profitability, an unprecedented development. Criminal activities, including theft and fraud, are occurring across industries, significantly affecting the retail sector’s profitability.
He further suggests that the advancement of technology, particularly the prevalence of social media and online selling platforms, has contributed to the sophistication of criminal activities. In the digital realm, identity theft has become more accessible as individuals can quickly obtain fraudulent identification online.
Chris points out that obtaining prosecution for these crimes has become increasingly challenging thanks to legislation that has greatly decreased the probability of being caught, prosecuted, and facing significant penalties.
As a result, there is a growing realization that theft, fraud, and other unlawful activities in retail environments target wealthy corporations and diminish the quality of life in local communities and cities.
Chris then brings our attention to two key factors when identifying fraudulent activities: repetition and the magnitude of the loss. He explains that behavior patterns, such as repeated return, loyalty, or credit card fraud, are important indicators to look out for.
While talking to Emerj, he points out that if anyone wants to stop fraud, they can do it overnight – but only with draconian measures that will negatively impact legitimate customers and sales. Therefore measures like link analysis plays a significant role in breaking cases, where technology and exception-based reporting assist in identifying connections between individuals involved in fraudulent activities.
By examining shared vehicles, addresses, or sightings, investigators can identify larger groups and prioritize security efforts accordingly. Chris acknowledges that dealing with vast amounts of data can be challenging, but extracting meaningful patterns and connections is essential for effective action.
Chris also suggests that criminals may utilize AI technology to assist them in their illegal activities, although he doesn’t have a specific example. He explains that these criminals gather data from the dark web, including personal information such as addresses and social details, which are readily available online.
He emphasizes that some individuals within a company, who are familiar with its policies and procedures, may share the internal information with their group of bad actors who then use the knowledge to defraud and take advantage of the organization.
Nelson further explains that they can identify anomalies and unusual patterns due to the wealth of available data and improved connectivity. If someone is internally stealing from the company, Chris and his peers are confident they will eventually be caught through exception-based reporting. They can spot outliers on the bell curve, such as abnormal returns, voids, or transactions.
However, Chris expresses concern about sophisticated individuals who possess inside knowledge and share information without directly engaging in illegal activities. These fraudsters create a more challenging situation as they understand how the system operates, the rules, and the exceptions made for loyal customers, making it potentially more difficult to detect their actions.
Harnessing AI and Data for Fraud Detection and Prevention
Chris Nelson discusses the importance of understanding the situation and clearly shows the challenges faced in fraud prevention. He explains that analyzing data and identifying patterns can differentiate fraudulent behavior from the norm.
Providing an example related to returns fraud, he acknowledges the common concern that restricting returns might impact a company’s best customers. However, he suggests looking at the situation by analyzing return patterns, including their frequency and dollar value.
By doing so, they can observe that the individuals they aim to stop are significantly deviating from the typical behavior of customers. These individuals represent a small percentage, often around 1% or less, of the total returns experienced by the company or store and are considered outliers in terms of their return activity.
Chris acknowledges that proposing to restrict returns may initially face resistance and skepticism. However, he highlights the importance of presenting the data-driven insights they have gathered.
By demonstrating that a small percentage of customers is responsible for a considerable portion, such as 20%, of the issue, the proposition becomes more acceptable. It then becomes evident that the goal is not to affect a significant percentage of customers but to address a specific outlier group.
Thus, Chris emphasizes the need to balance security measures without being overly restrictive, considering the impact on commerce. He notes that the approach to security has evolved, moving from physical security to data security and now to virtual security. Data analysis plays a crucial role in directing their efforts.
Chris expresses excitement about incorporating AI into their security systems. AI can help detect anomalies, such as through AI-powered video surveillance systems, which can alert operators to potential threats. Further, combining AI with data analysis and link analysis can improve their ability to identify exceptions and enhance fraud prevention efforts. Chris finds the development in technology and data-driven approaches highly promising.
However, the move could pose a question about the privacy of employees. As a solution, Danya Golan, Chief Marketing Officer at Hailo, suggests computing the video on edge.
“Stores create an enormous amount of video streams, from full HD cameras, with 30 frames per second, 24 hours a day, sending a lot of video data into the cloud for analysis – that’s a lot of bandwidth and a lot of costs,” she tells Emerj. “And it’s a hazard to people’s privacy – employees and customers alike. When the video analytics are done on the edge, the video input is extracted and summarized, and only the very important insights or inputs are transmitted and recorded in the cloud for future analysis purposes.”
“AI can be used to identify abnormal behavior and analyze people’s pose to recognize suspicious behavior, creating a real-time alert to security personnel. AI can also be used to classify buyers by age, gender, etc., for statistical purposes, which could be used later to improve customer satisfaction and provide a more personalized shopping experience. But you don’t have to jeopardize anyone’s privacy or personal identity to perform all these tasks, and you don’t have to stream this huge amount of data into the cloud.”
– Hailo Chief Marketing Officer Danya Golan
Lastly, Chris emphasizes the significance of data as a valuable resource. He mentions traditional exception reporting, which allows him to identify specific patterns or anomalies by manually interrogating the data. However, he acknowledges that the volume of data in the virtual world is vast, making it challenging for a human to analyze it comprehensively. This is where AI comes into play.
Chris expresses the need for AI to uncover patterns and connections that may not be immediately apparent to human analysts due to the sheer amount of data. He highlights the potential of AI to reveal commonalities among fraudulent digital transactions, providing insights that can aid in fraud detection.
Chris also shares an example of store shortages, expressing curiosity about the indicators that precede such occurrences. He explains that while inventory management can identify missing items, AI can help identify other key performance indicators (KPIs) that correlate with high shortage rates.
He concludes by telling Emerj that AI may not provide definitive answers as the “smoking gun,” but it can guide decision-making and point the team in the right direction.