The Analytic Steroids for Digital Lending

By: Amit Das, Founder and CEO of Think Analytics and Former PwC analytics expert

Lending in India is growing exponentially. SME Lending will be more than $500 billion, while consumer lending opportunity is expected to $1.2trillion+ by 2020. Today, only about 1-2% of this lending market is “digital”. But that’s a fact from yesterday. But, if we were to look at the historical growth in digital lending, we would completely miss the wave. In the next 3-5 years, Digital is expected to contribute more than 15-20% of the lending books.

To get ready for this growth and scale – lending would need to be on steroids. They would need to harness the latent potential of alternative data and smart processes.

Why does alternative data matter?

Let us say your company’s approval/ disbursal rates are in the 10% range. For every loan application that you evaluate, you are paying Aadhar/ PAN verification/ Bureau charges, Field Investigation fees, additional processing/ documentation expenses, and even the most optimized organizations end up spending ~Rs.1000+ per approved loan application. Maybe more!  This makes decision making cumbersome, and lending expensive. Now, if there were a smarter way of assessing only those applications that need to be really verified against these external agencies, one could bring down one of the component costs of lending. And this is just the first part of the value chain of lending that alternative data can change.

Today, businesses have access to a large amount of alternative data the moment a user downloads their app. Digital world provides seamlessly authenticated integrations with agencies – UIDAI, NSDL, Bureaus, etc., access to alternative data-  messages, locations, app usage, call patterns, etc. as well as a user’s behavioral profile – likes and interests, social relationships, etc.

What kind of alternative data is being leveraged today?

The utility bills and recharges history can be used to triangulate how vested you are with your current place of residence, and whether you have stable financial behavior. Integrating with your Uber trips history and driver ratings, or noticing how often you return the items you order on e-comm sites can provide additional insights into your behavior. The old world philosophy of “introducers”/ “guarantors” can be replaced with a “Digital Vouch” network. The thematic bottom-line is – You are what you do. And in the digital world, there’s a record of almost everything you do.

A similar approach applies to the organizations as well. As a simple example, on the Aliexpress app, one of the strongest filters a buyer looks for is the seller ratings before one buys something. We can augment the seller ratings with a NLP and Machine Learning based solution for interpreting the text feedback to normalize the ratings. Such algorithms further de-clutter the cultural biases that feedback givers may have. Similarly, in India, by reviewing the JustDial ratings, consumer complaints against a business, the depth and breadth of social presence (such as facebook likes), ROC filings could emerge as important parameters. Algorithmically evolving, processing financial statements, GST filings, EPFO records, tax returns, etc can provide critical insights into the health of the business – from long term financial parameters as well as short term cash flow perspective.

Are businesses ready to leverage alternative data?

We expect financial institutions to go through a 9-12 month cycle before alternative data is ubiquitous in their decision engines. Being said that – The winners will be those organizations that have started building those experiment engines today. Over the last several months, since we have gone live with Algo360, we have been involved in several path-breaking conversations with banks, NBFCs, new age fintechs and these conversations paint a clear picture of tomorrow – where a significant part of lending will be truly digital with minimal physical interventions.

Leave a Reply

Your email address will not be published. Required fields are marked *