We would like to wish you and yours a safe and happy holiday season! While many people are taking time away from the office this week, still plenty of others – especially at financial institutions – continue to punch their proverbial time cards. Another hot topic in our industry that seemingly never takes a day off is fraud.
As holiday spending ramps up toward the end of the year, more and more crooks out there are scheming ways to get something for nothing. Credit fraud continues to be a growing concern in the lending industry. Financial institutions need to find ways to increase fraud detection and prevention capabilities within their automated decisioning technology and processes. Analytics continues to provide a trusted and viable counterpunch to rising threats of fraud.
Predictive variables are often used in the development of scorecards based on multiple information sources. The following types of variables are often analyzed to provide an assessment of fraud:
- Standard credit bureau variables: The main information source for these variables is usually the credit bureau file. These variables are typically used for (and are also more oriented to) credit risk. For fraud prediction purposes, some variables offer decent predictive ability but less effectiveness than customized fraud variables. Some examples of predictive credit attributes include the number of inquiries during different periods of time, age of the accounts, age of the credit bureau file and the number of satisfactory accounts over different periods of time.
- Custom fraud variables: The main sources of information for these variables are often the credit bureau file and application. These variables are sometimes the strongest predictors of all the different types of variables calculated. Unlike standard attributes built by risk teams to predict risk, custom fraud attributes built with the intention of predicting fraud tend to be more effective. Some examples included address stability, inquiries based on more sensitive and specific periods of time, and inquiries of different contract types.
- Geocoding variables: Geolocation variables are calculated from mapping software based on precise locations. Strong fraud predictors include calculations of distances and locality matching flags based on client home address, dealer address and employment addresses on the credit bureau file.
- Zip code variables: These variables are calculated to assess the fraud prediction by geographic area (zip code). Zip code variables are sometimes weaker than the same variables at individual levels but still useful for building area-level profiles to compare to the applicant information. They identify patterns of credit activity in specific areas that demonstrate fraud characteristics. For example, specific types of credit recently opened in a particular area.
To download a copy of this complimentary eBook, please click the button below.
Photo Credit: CafeCredit.com
