The challenge | Enhancing fraud detection, scalability and efficiency
FanDuel, a prominent player in the gaming and sports betting industry, faced a critical challenge. It needed to simultaneously reduce suspected fraud while speeding up the detection and prevention of fraudulent activities - all of which would involve automating its fraud detection processes. One of the primary hurdles was scalability, as FanDuel’s existing systems struggled to keep up with the sheer volume of transactions on its platform. Equally important was the need for effective tools to swiftly evaluate these transactions. Credit card fraud emerged as the most frequent threat, instigated by perpetrators of fraud who illicitly acquired credit card details and used them to place fraudulent bets on the FanDuel platform. Geo-fraud, another significant concern, involved users attempting to manipulate or bypass geolocation restrictions imposed by sports betting platforms. Since online sports betting operates under state-level regulations in the United States, users could only place bets when physically present in a state where sports betting was legal. FanDuel's challenge was compounded by the absence of an automated fraud detection system. Instead, it relied on manual processes. Reports were generated from various systems and data stores, and its fraud analysts painstakingly compiled and processed the data, often using spreadsheets. This labour-intensive approach enabled investigations to be carried out but introduced a critical one-day delay in the process. The one-day delay had substantial repercussions, affecting both the business and customer experience. In an industry with several competing gaming platforms, any delay posed a risk of losing customers. For genuine customers, the one-day delay might lead them to explore alternative platforms, resulting in potential customer loss for FanDuel. From a fraud prevention perspective, the one-day delay hindered capturing and blocking fraudulent users before they could seek alternative methods to engage in fraudulent activities.
The solution | Advanced fraud detection and prevention
In response to FanDuel's multifaceted fraud detection challenge, xDesign introduced a third-party vendor solution renowned for its cost-effectiveness and superior accuracy in identifying fraudulent activities. This innovative solution harnesses the power of machine learning to scrutinise user behaviours and activities, crafting detailed user profiles critical for evaluating potential fraud risks. The solution revolves around real-time scoring and risk assessment, utilising user profiles and behavioural patterns to assign numerical scores to transactions. Transactions exceeding predefined thresholds are promptly flagged as potentially fraudulent or high-risk, enabling rapid intervention. To bolster fraud detection accuracy, the solution leverages a vast client network, combining FanDuel's internal data with global client insights via proprietary APIs and data models. The solution embraces an event-driven architecture, seamlessly integrating with FanDuel's existing systems. Key components include Lambda functions (small programs that run automatically to perform tasks) for data orchestration and critical processes, SQS (Simple Queue Service - effectively a waiting line for tasks) as a resilient event repository, and SNS (Simple Notification Service - a system for sending updates or alerts) for seamless event distribution. An API Gateway serves as a secure entrance where analysts can manually make decisions about potential fraud. Designed to ensure only authorised personnel can access it using a specialist software application, it uses a list of approved IP addresses and robust verification methods to keep the process secure. Manual analyst decisions serve to continually train the machine learning model, improving its accuracy over time. These decisions get processed through the same solution for automated events via SNS to ensure consistency between both sources.
The results | Reducing fraud investigation times from one day to minutes
FanDuel's new system brought about significant improvements in fraud prevention. The time needed to investigate potential fraud plummeted from a day to just minutes. The new system is fully serverless and offers automatic scalability, allowing FanDuel to handle large events like the Super Bowl without the need for extensive configuration changes or bills for redundant infrastructure. This scalability, combined with better data analytics and investigation tools, has expanded the system's capabilities. One of the annual targets was to reduce total client fraud by 10%, and FanDuel was running at an 86% success detection rate based on a recent report. This level of success potentially translates to saving clients $1 million a month. Aside from these direct benefits, there have been several other positive outcomes. The fraud prevention team now has more time to investigate other areas of fraud, making their efforts more comprehensive. The project has also increased the importance of the fraud prevention area, leading to recommendations for integrating additional security measures. FanDuel intends to integrate more vendor offerings into the system to prevent various forms of fraud in the future. The focus is shifting towards preventing fraud before it occurs, such as checking deposits before they're made. The roadmap is ambitious, with a larger scope and more teams involved. To adapt to emerging fraud types and changing behaviours, the system relies on refreshed vendor data models and the addition of new vendor tools and offerings. These changes help fill any gaps and enhance fraud prevention workflows and data analysis.