Financial Fraud Indicators Revealed in Common Data Collection

The value of fraud loss for the credit card industry tops out at $131 million, according to a 2018 United States Federal Trade Commission Report on Payment Fraud. The perpetrators of financial fraud have improved the sophistication of their methodologies while the industry has simultaneously grown with an increasing number of payment processing providers. 

The news headlines are replete with companies learning about or disclosing fraud after the fact. This begs the question: how valuable would it be to detect and act on fraud when it’s occurring? We believe that real-time analysis of available data can shed light on areas of high fraud, by evolving the methodologies in the system to capture it before the fraud happens.

The different ways in which enterprises detect anomalies are exhaustive, but in our recent work with a financial services client, we’ve identified three no-fail approaches that can help fraud strategists analyze and re-design the way they capture real-time fraud data.

  1. Analyzing a rule-based data set
  2. Analyzing outputs based on an in-house training model
  3. Identifying outliers using a third-party scoring system

Our client was capturing data and information based on different events; however, the organization ran into trouble when maximizing data utilization. Further, analytics, visualizations, and highlights were not accessible to analysts or executives to help visualize patterns and identify ways to improve detection and prevention methodologies. Borrowing on learnings from this project, how can analytics teams apply similar principles to enable a fraud detection system that’s able to keep pace with a continuously evolving threat?

  • Align to business needs. Work with key stakeholders to identify business needs and build use cases to translate the requirements into technical terms. In our project, we interviewed subject matter experts to identify different data elements and their relationships.
  • Build an adaptable model. Focus on building a flexible data model that can be repurposed for a varied target audience. We designed a NoSQL-friendly data model that could efficiently tie different events based-data. Concurrently, we also built a product with custom visualizations using Tableau. This approach provides visibility into identifying the patterns in the rule-based data, outliers to the scoring patterns, and other event-based data elements that help analysts to quickly notice the problem area and re-strategize their existing methodologies to capture fraud. Ultimately, this helps customers and analysts take better actions and decisions to mitigate fraud as early as possible.

Data and analytics—powered by visualization tools like Tableau—equip fraud strategists with real-time insight into their rule-based data, scoring patterns, and gap analysis on transactions. These are key to helping organizations to quickly identify problem areas and re-strategize existing methodologies to capture fraud—informing smarter actions before it’s too late. Taking this approach arms businesses with the unprecedented opportunity to visualize insights that curb the issue at a near-atomic level, while at the same time enabling a high level of customer experience.