How It Works

A step-by-step guide to Monsoon's AI Pipeline

Steps to lower delinquencies and higher approvals

Steps to lower delinquencies and higher approvals

Prediction

Data Collection

Feature Engineering

Model Building

Prediction

1.

Data Collection

Data Collection

Data such as bureau data, repayment data, financial data, form data etc are collected from the customers. These data points are thoroughly analyzed for accuracy and validated for any errors before being used for building models. We work on the data collected by the organisation and do not bring any new data on board.

2.

Feature Engineering

Feature Engineering

This is our secret sauce. Our Intellectual Property that we have developed on our own over the course of 5 years. You would not find these on any open source platforms as we have meticulously developed the pipeline over the course of our lifetime. Features based on domain knowledge, expertise and also complex statistical patterns as well as a combination of both are generated and used.

3.

Model Building

Model Building

Monsoon uses state-of-the-art feature selection techniques, some of which are widely used by the best-in-class data science driven organizations and other that are proprietary & have been built in-house. Feature selection is a critical process that enables models to perform better in the real world (production environment) by:

 

  • Minimizing the chances of overfitting i.e. of learning patterns present only in the training sample & not in the real world
  • Improving model stability & reducing variance in performance
  • Making models more robust to fluctuations in distributions of data seen in the production environment. 

4.

Prediction

Prediction

Model scores will in-turn be validated by the client and recommendations compared with the actual outcomes (on the OOT set) to measure the performance of models on a regular basis.

It is anticipated that our technology will substantially reduce the delinquency rate & increase approval rates  while identifying signs of stress much earlier than existing models.

The actual results of this phase and the key metrics associated with the portfolio such as the following, will be shared by the client with Monsoon:

  •         Delinquency rates
  •         ROA/ Credit loss adjusted NII

By using Monsoon's proprietary AI technology, you can

Delinquencies by 30%    Approvals by 25%    Turn Around Time by 40%

Revenue by 35%    Credit Losses by 22%    Value At Risk by 33%

Delinquencies by 30%
Approvals by 25%
Turn Around Time by 40%
Revenue by 35%
Credit Losses by 22%
Value At Risk by 33%