Customer relations are challenging for a bank from the moment they begin. However, the more products the client uses, the closer the relationship gets. Our client's goal was to create more sales opportunities among existing clients.
To build profits, banks need to be more effective at cross- and up-selling to existing clients. This calls for subtlety: an overly aggressive approach pushes customers away, while repetitive and inappropriate offers also lower customer satisfaction.
The lack of proper tools for selling additional services to existing clients hampers profits due to the bank missing opportunities.
Our solution predicted the probability of each customer taking out a card. After testing multiple approaches, the team produced a boosted tree-based solution. The model analyzed a number of variables, including customers’ historical data about the products they used and their transaction history. With these, it predicted how willing they would be to open a credit card account.
To address the client’s concerns about the performance of its database, we reduced the model’s size to only 20 variables (out of over 300). This had only a negligible effect on the accuracy.
Using decision trees and XGBoost helped deepsense.ai’s solution outperform the bank’s current solution by 12.5%.