Statistical issues in retail banking Seminar

The personal or retail banking sector presents a wide variety of stimulating challenges for statisticians. In an increasingly competitive environment, with more and more financial products available, the edge provided by a superior statistical model can make the difference between success and failure. Models are needed to understand how potential customers may behave, how existing customers will behave, and how they might behave if interventions are made. Particular problems include decisions on which customers to accept, what interest rates to charge, monitoring behaviour over time, and detecting fraud. Often the decisions must be made rapidly. The data sets can be extremely large, they are constantly growing, the underlying distributions evolve over time, they may be high dimensional, involving mixed variable types, and the data quality is often surprisingly poor, suffering especially from missing values, selectivity bias, and data degradation. This talk presents a high level view, illustrating some of the challenges and solutions.