Using fraud scoring to improve first party fraud detection
RBS wanted to take a fresh look at its fraud scoring models and improve their predictive power. Knowing that Jaywing’s analytical capabilities in fraud were second to none, they turned to us for help.
We were commissioned to support the internal development of a customer level model to assess the risk of first party fraud. This model was to be used for assessing new applications for credit and pro-active management of high risk customers.
Throughout a year-long project, working mostly on-site and consulting with the internal team, we provided end-to-end support from sample design to review of internal development documentation, culminating with delivery of an audit report.
We identified two clearly distinct types of first party fraud that meant building separate models for each to ensure better fraud detection.
Rosemary Byde, Risk Infrastructure & Analytics, Predictive Analytics, Retail Wealth & Ulster, said: "For this project, the refinement of the bad definition led to the development of more powerful and useful models.
However, for us, one of the most successful aspects of the project was the way that Jaywing worked with us to develop the bad definitions. They provided the experience, insight and drive to investigate further, but because this was done collaboratively, we have acquired a thorough understanding of the end result.
In this case, we are better able to explain the differences between the two models and to advise on where they might be used most effectively as different scores are appropriate for different kinds of strategies."
Let's work together
Talk to us