Fraud Model – A tool designed to help segment your population and determine the likelihood of fraud. The customer logs onto www.yourhomepage.com and says she is looking to open an account. By asking her for basic personally identifiable details, a fraud solution can ask a lot of questions. Am I getting an SSN? Is this a real SSN? Can I link this SSN, name, and address to some externally verified data source? Is there anything suspicious about this transaction? How many times have I seen this SSN recently? Is this the third time in a week I have seen it through online channels? Has this SSN recently been reported as a stolen identity? The earlier you introduce a fraud model into your loan origination process, the quicker these red flags can be set off.
Half of those questions indicate fraud that no bank should entertain. If too many of the questions have the wrong answer, stop processing! Don’t try to extend an offer of credit, don’t consider cross-sell, and certainly don’t waste time and money asking her an invasive set of questions she won’t be able to answer. If a potential client customer can’t give you a name and social security match, is she going to know who provides her car insurance?
The data produced by a risk model can speak volumes to a risk analyst. Ultimately your risk analysts are paid too much to handle these credit requests individually and should only be managing process. Rule flows should determine how every question answered by the model impacts the credit request. Is it a real SSN with no history of abuse but the address is a prison? Perhaps this request needs to flow to an authentication solution. Is this a valid identity but the SSN has been used four times through your online channels today? Maybe this should flow to a call center representative. A good fraud score and reason code driven process creates the most information early on in a request’s lifecycle, allowing for the most tangible benefit later.