financial services

Business problem

A large commercial bank needed a new "mass affluent" (based on investable assets) targeting model to identify growth prospects for its wealth management business unit while better utilizing its own information and relying less on third party data.   The Bank required the new solution to be created, tested and deployed in less than 90 days. Concurrently, the Bank’s existing analytic team needed to be trained on new analysis software.

Solution

Parallel sub- projects were established to define business user needs and requirements; assess, request and assemble data to be used in the analysis; assemble a multi-terabyte analytical environment on-site using conventional and advanced genetic algorithms analytic software; and train the analytic users in the use of genetic algorithms.

Careful analysis of business user needs and requirements revealed that multiple models were required to meet the Bank's objectives.  First was to establish which customers were likely to meet the Bank’s “mass affluent” definition.  Just as important to the bankers was the ability to predict which customers were likely to increase their deposits, investments and balances after becoming wealth management customers.  Successfully aligning these two overlapping requirements was fundamental to the project’s success.

Use of genetic algorithms enabled the Bank to explore potential predictive data in ways never before possible.  Hundreds of variables, and their interactions, could be evaluated.  Consequently, the resulting models could be based more on the Bank’s proprietary information and less on third party data.

Simultaneous to model development, the Bank’s analytic team was trained on the GA software and deployed in sub-teams to successfully produce additional cross sell and retention models.

Results

The new mass affluent model was successfully created, developed and deployed in less than 90 days.  The new model correctly predicted 90% more mass affluent households than the Bank’s existing model, while the growth model identified prospects most likely to grow.

The Bank's analytic team successfully adopted the new genetic algorithm software and was able to increase the number and complexity of models delivered for the business.