Organizations commonly depend on predictive models to help mitigate fraud losses. Custom analytic models incorporate client data and performance information to provide the optimal analytic scoring products and services for application fraud detection. A key to success in developing custom analytics is having accurate and predictive data to use for model development. This requires a representative sample of good (legitimate) transactions and bad (fraudulent) transactions, with accurate and holistically populated input data. It is important that the fields and filters used to define the modeling data set properly represent what will eventually be available in a production system.
Another critical component of successful model development is to have well-defined performance tags to properly train the model. This often can prove problematic, as first-party fraud and third-party fraud exhibit different behaviors and can be difficult to clearly and consistently identify. In addition, bad credit behavior often can be confused with fraudulent behavior and add to the "noise" of the performance tags.
The Fraud and Identity Solutions Consulting and Analytics team provides an analysis service that helps clear up this confusion. Experian has the benefit of multiple proven production scores tailored for credit, first-party fraud, third-party fraud and combinations thereof. By using this combination of scores and a deep understanding of how certain features perform on certain types of losses,our team is able to provide a clear understanding of the underlying loss behavior. This, in turn, can be used to help fine-tune an appropriate tag to aid in custom model investigation and development. This analysis service has been used successfully at a variety of clients ranging from banking to retail to payday loans.
For more information on Experian's forensic fraud loss analysis services, click here.
Contact your Experian account executive or call 1 888 414 1120 for more information.