Churn Prediction churn

USECASE/DOMAIN: Existing models cannot accurately predict customer churn when reason for attrition is not captured.

FIBS MODEL: We used tree based models for non-time series data such as demographics information. Along with that we used neural networks for time series data such as activity and behavior of a customer.

Using customer demographic, bill payment history and balance management, support tickets, call and data usage of customers (where available), we can predict which customers are likely to leave the service.

Customer Risk Profile searchrisk

USECASE/DOMAIN: Identify customer risk profiles for better pricing of the insurance products.

FIBS MODEL: We processed existing customer datasets to learn the patterns of the people. Based on the learning, our tree based model continuously identifies profiles of the customers from High Risk to Low Risk.

We can segment customers using their demographic and behavior data for:

  • Identifying new customer risk assessment
  • Credit limit enhancement
  • Targeted marketing and promotion
  • Increased retention

Unlike traditional models, AI learns continuously. This ensures the model becomes more accurate as new data is fed and does not become obsolete.

Customer Segmentation and Targeting nnet

USECASE/DOMAIN: Conversion Rates of the bank for short term loans were low, leading to reduced sales. It was difficult for the company to target eligible customers with buying intent.We can segment the customers and recommend products and services for existing customers by using customer demographic details, bill payment/ balance history, support tickets, and call and data usage (where available) of customers.

FIBS MODEL: We used Deep Convolutional Neural Network on 18 months of data for existing customers including demographics, history & activity on bank accounts and credit scoring to predict customers most likely to purchase.