FREE DM Review Site Registration!
Sign-up today and access DM Review on the Web!

Your FREE registration entitles you to:
FREE email newsletters

FREE access to all DM Review content

FREE access to web seminars, resource portals, our white paper library and more!

   

Publisher reserves the right to serve qualified requesters only.

The Importance of Fraud Analytics

BI Analyst Take

Fraud's effect on an organization's bottom line is just the tip of the iceberg. Without a proactive approach to combating fraud, the ability to gain and to maintain customer loyalty is almost non-existent. According to Meridian Research, in 2005 alone, the cost of U.S. companies not using anti-fraud applications was $60 billion. Additionally, organizations are said to lose an average of six percent of their annual revenue to fraud and abuse committed by internal employees. An organization's loss of revenue is only the immediate issue involved in not dealing with fraud proactively. Lack of customer faith and of perceived security lead to long-term loss in revenue and the inability to stay competitive in a quickly changing market.

A proactive approach to fraud identification is the only way to address and to lessen the effect of fraud on organizations today. Aside from the security provided to customers, the amount of money saved by organizations is large considering the financial payoff of implementing a fraud analytics solution. This column addresses financial-based fraud; other related topics, such as the use of analytics for homeland security, etc., will be discussed in a subsequent acolumn. This column also identifies how fraud analytics differs from general analytics use within business intlligence (BI), the human factor required for effective fraud detection and the technical implications of implementing fraud analytics within an organization's current business intelligence environment.

The Use of Fraud Analytics within BI

Organizations with mature business intelligence environments can integrate fraud analytics within their current environment to take advantage of processes and architecture that are already in place. This approach may be easier than starting from scratch. Although the applications are different, the same infrastructure allows multiple types of analysis to be conducted, to be reported on and to be acted upon within a BI-based framework.

BI applications offer organizations a way to detect potential fraudulent activities before they occur. Data warehouses collect financial-based information and create what-if scenarios to identify how external factors and market changes affect sales, product mix and operations. These same technologies can be used to gather information and use the same type of predictive analytics techniques to identify suspicious patterns. The main difference between regular use of BI and fraud-based applications is the way information is collected and analyzed.

The identification of potential fraud involves the opposite approach of traditional data warehousing. Developing BI solutions involves taking a subset of transactional data to identify trends, to measure current performance against metrics, to identify the success of marketing campaigns and to measure sales. Instead of identifying subsets of data, gathering and storing all potentially relevant information to mine that data to find patterns of usage and discrepancies to classify potential fraudulent activity becomes important. This requires the collection of information related to people's interactions and associations with one another, commonalities among submitted claims information or address and name data on Social Security documents to identify suspicious activities and overlaps of submitted data. Instead of developing predictive analytics models based on what-if scenarios, flags are created due to statistical probabilities to identify overlaps of information or patterns that identify when statistical probabilities have been reached or exceeded. This allows organizations to manage potential threats before they occur as well as identify patterns within data that may not have been obvious beforehand.

Human Intervention for Fraud Detection

The responsibility to combat fraud lies on the shoulders of organizations. Although consumers can take precautions to protect themselves against fraud, organizations need to provide consumers with a sense of safety. The costs of not doing so can be measured in loss from bad debt and breaches in security. Through the use of fraud analytics, organizations can identify suspicious behavior and patterns before fraudulent activities occur. General analytics are designed to find patterns within data that would not be easily recognizable by people. Within fraud analytics, the same holds true, however the patterns recognized to identify potential fraudulent behavior represents the beginning and not the end of the analytics process.

The main difference between the use of fraud analytics and other applications of analytics is methodology. By implementing a solution to combat fraud, organizations are taking the first step toward a proactive approach. A methodology includes how employees handle flags and react to stop fraudulent activities, either before they occur or early on within the process, thereby lessening potential threats in the future.

The application of fraud requires human intervention to identify whether there is a basis to pursue the recognized pattern. This can be seen in the automotive industry where departments use analytics to determine whether dealers are submitting a suspicious amount of warranty claims or within insurance companies that identify like information on claims to determine an inordinate amount of overlaps. In both of these examples, employees are responsible for identifying and managing fraud. Without employee buy-in and an added importance placed on the human element, organizations will probably not develop a sustainable fraud detection solution.

Can the Current Environment Handle Fraud Analytics Data?

The implementation of fraud analytics within the organization requires a different data warehousing structure. If organizations choose to implement fraud analytics within their current BI infrastructure, they should evaluate the current environment. Even though BI may be mature within an organization, the use of fraud analytics generally requires a more robust technical architecture as more information is being collected. The information collected includes areas such as trend identification and linking different data sets that include unstructured data sources as well (for example, the information stored in email or documents). This means that current data warehousing environments may not be robust enough to handle the amount of data being collected, hence the use of data warehouse appliances or databases that can handle multiple terabytes of data.

The threat of fraud should be enough for organizations to deploy fraud analytics. The key areas to identify are the differences between fraud analytics and the current applications of business intelligence, the human element required to optimize the anti-fraud solution, and the technical considerations. Armed with this understanding, organizations can move from a reactive environment that responds to fraud once it has occurred towards a proactive fraud detection environment that identifies potentially threatening actions before they occur.


Lyndsay Wise is an industry analyst for business intelligence. For more than seven years, she has assisted clients in business systems analysis, software selection and implementation of enterprise applications. Wise also conducts research of leading technologies, products and vendors in business intelligence, marketing performance management, master data management, and unstructured data. Check out her blog at myblog.wiseanalytics.com.

For more information on related topics, visit the following channels:



Industry Vendors