Monsoon’s core offerings
Machine learning powered predictive models
Monsoon’s machine learning powered models identify patterns that link customer data (e.g. credit bureau data, bank statements, demographics, location related data and if available, alternate data) with outcome data (e.g. repayments, response to a product offer etc.). These patterns tend to persist over time and can therefore, be used to predict a future outcomes given customer data (non-PII data). Monsoon’s machine learning models “trained” on past data are able to effective predict outcomes given customer data and enable lenders to unlock value by reducing future delinquency rates, improving cross-sell revenue, reducing collections costs and making offers to customers that tend to get accepted.
Further, these patterns are unique to every lender based on the way they source customers, underwrite to borrowers and collect from borrowers. Our machine learning engine “learns” these patterns from every lender’s historic portfolio and application level data, combined with the repayment data associated with every borrower in the portfolio.
Core offerings
Scorecards for Banks, NBFCs, HFCs and Fintech lenders
Application scorecards
Use data such as Bureau data, Banking data and/or alternate data to enable quick approval of safe loan applications and rejection of risky ones.
Reduce delinquency rates by up to 30% and increase approval rates by up to 25%.
Behavioral Scorecards
Use Bureau data, transaction data and past repayment behavior to predict the risk of existing loans on book, enabling timely intervention.
Reduce risk-cost by up to 35% and opens up cross-sell opportunities.
Collection scorecards
Use bureau data, past repayment data and responses to collection efforts to predict which customers will miss the next payment.
Enable the prioritization of collection-efforts & up to 40% reduction in collection costs.
Pricing Models
Risk based pricing models enable you to price loans (interest rates, LTV, tenures on the basis of customer risk) and pricing power.
Optimize ROI, tenures and other loan parameters on the basis of borrower risk.
Collection scorecards
Use bureau data, past repayment data and responses to collection efforts to predict which customers will miss the next payment.
Enable the prioritization of collection-efforts & up to 40% reduction in collection costs.
Pricing Models
Risk based pricing models enable you to price loans (interest rates, LTV, tenures on the basis of customer risk) and pricing power.
Optimize ROI, tenures and other loan parameters on the basis of borrower risk.
Cross Sell scorecards
Leverage data available with the lender to predict which customers are likely to accept cross-sell offers & for which products.
Drive up to 35% more cross-sell revenue.
Customer Attrition Models
Predict which customers are likely to close their loan/deposit accounts, months before they actually do, enabling intervention.
Help reduce customer attrition by up to 24%.
pre-approval models
Help raise book-size with minimal risk by pre-approving customers from liability base for asset side products
Help approve up to 35% of the base at minimal risk.
Early Warning Models
Similar to behavioral scorecards but also account for infant-delinquency i.e. delinquency within the first 1-2 months of disbursal.
Help cut risk-cost similar to behavioral scorecards.
income estimation models
Enable the quick-estimation of income based on bureau data to help lenders ascertain the true income of a customer. Used during underwriting.
Work in tandem with application scorecards.
Pre-pd models
Enable quick risk-bucketing of applications before expensive Personal Discussions/ Field Investigations so that low-risk customers can be exempted.
Help reduce costs related to field interviews by up to 40%
Alternate data based scorecards
All the models and scorecards that are mentioned above can be built with alternate data such as SMSs, GST returns, Fastag transactions, UPI transactions etc.
policy consulting
In addition to models and scorecards, our team helps you optimize policy based on data, to enable you to reduce risk while making full use of new opportunities that the fast-evolving environment has to offer.
Thoth- A No-code ML Modeling and Advanced
Analytics Suite for the Business User
Intuitive UI Wrapped Around Monsoon’s Powerful ML Pipelines
Data
Data Analysis
Modeling Strategy
Feature Engineering
Feature Selection
Modeling
Validation
Models
(Downloadable)
Commercial Models
Predictions
as a Service
Bespoke Commercial
Arrangements
Subscriptions to
Thoth
Predictions as a service
We expose a range of easy-to-integrate APIs hosted on all 3 major Cloud Providers. As a customer, you only pay for every prediction that is generated by an API call. No development fees, no cloud bills and no need to deal with ensuring that the scorecards are up-to-date.
Bespoke Arrangement
Bespoke commercial arrangements in which models are built for you and the deployment code is shared with you are also on the table. These contracts typically work on an annual basis and cover a fixed set of predictive machine learning models including development and maintenance.
Subscriptions to Thoth
A subscription to THOTH, enables you to build enterprise class models within 4-6 hours- a process that normally take weeks or months. This allows you to experiment fast, without the need to get a litany of approvals for every model, not to mention the cost of hiring so many Data Scientists to do this.