Calibrating a Scoring Model into a Decision Model in Banking

To make Analytics models give real-time recommendations with accuracy is an important part of creating Automated Decision Systems in organizations. As mentioned, at Pexitics we focus on Intelligence amplification and helping teams make great business decisions, every day.

This article deals with customizing the ItP Q12 — Intention to Pay model from Pexitics for Retail Lending. In the earlier article I had discussed How Psychometric ‘Intention to Pay’ scores should be implemented (

Today’s article is a continuation of the process -and I will walk you through how the Intention to Pay model is customized for decision making for each lending organization. Infact, it is customized for each product, location and demographic of customers for the NBFC / MFI

The Intention to Pay Q12 is a 12-question psychometric assessment which evaluates a customer as a person and scores him / her on how they will react to situations of financial stress. It helps Retail Finance NBFCs and MFI to improve their customer understanding at 2 levels:

1. In situations of financial stress, how will a customer react (suicide, get aggressive, be responsible and work out a mutually beneficial plan)

2. At point of collections, how can collector tailor their communications to ensure maximum recovery

There are other benefits, like customer portfolio familiarization for a new Relationship manager.

The ItP scores use Random Forest at the backend to create micro-segments and map them to 3 decisions: Proceed, Caution, Avoid. This requires that the NBFC / MFI runs a pilot of the ItP Q12 with its best and worst existing customers for the system to map out the micro-segments which should be Avoided. The Caution section should lead to a stronger scrutiny by the credit team while the Proceed decision should allow for fast approvals (coupled with Ability to pay scores) . Ofcourse, the assumption is that the Ability to Pay is on track — and usually, every lending organization has robust Ability based Credit Policies.

Thus how does the Loan approval decision matrix look after ItP scores are incorporated?

Sample matrix — ItP Q12 for Credit decision support

Thus ItP Q12 recommendations will help the credit team modify the offer in terms of price, loan amount and tenure — to better manage Portfolio Risk. If the option of taking collaterals is available, then this can help hedge the Risk through collaterals.

This ItP Q12 Recommendations will vary with each portfolio — just like in some banks Personal Loan is approved at CIBIL 650 but for other banks the minimum CIBIL score requirement is 700. The calibration of the ItP model scores therefore becomes a very important part of the process. The methodology is simple: run the ItP Q12 for 100–300 existing customers, both with good repayment track and delinquent track for FREE. This is a great way for your credit / customer relationship managers to get familiar with the product and for us to analyze and calibrate the Avoid/ Caution / Proceed recommendations for your portfolio.

The other way of course is to go with the Peer benchmarking concept- where the recommendations can be given as per statistical segmentation Eg: Bottom 25% — Avoid; Middle 50% — Caution and Top 25% — Proceed.

The expected outcome of this scoring model implementation is a more robust Credit Risk Model.

Improved Credit decision model

We look forward to being a part of your Risk and Profitability solutions. Do reach us for a discussion at .



Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store