Lending: The game of zero-sum and the quintessential challenge of risk prediction
Reduce the probability of defaults by accurately predicting credit risk almost in real-time for lending applications coming in bulk. The key is to augment risk prediction models with real-world location data.
TL;DR — Lending is a risky enterprise. To be at its best profit-making potential, businesses need borrowers with maximum tenure without the risk of defaults. The insights that banks and NBFCs work with are limited and do not effectively alleviate risk. Location intelligence helps measure ‘affluence’ — the intangible benefits of functional living or operating in a neighbourhood — which is directly correlated to default risk.
Why does the lending industry still experience loan defaults? The ecosystem is not that troubling to understand when put this simply — after all, what you as a bank or NBFC do is attract a large pool of customers to get an eventual hold on a niche set within them.
Although it is a little more complicated than that, let’s take personal loans, for example. Money is made on borrowers who should take the longest time to pay off their loans, but would, almost a hundred percent, pay you back. With this though, we begin edging towards the riskier fringe of lending which is where defaults are, about 12.7% of all loans in FY 2021. Borrowers who should take the longest, would pay you back, and must not default?
In retrospect, a bunch of analytical discoveries (such as total line utilization rate, months since the last public record, general compliance, etc.) can be made to describe who defaulters are and why they default. That still leaves room to effectively understand how an incoming customer is going to behave. It is an industry-wide problem — no lending business has that understanding of risk prediction available, considering that banks wrote off 2.02 lakh crores in 2021. What we are saying is, that we understand lending is a risky enterprise, given that borrowers come in all shapes and sizes.
Here’s what we find really interesting. What if these ‘shapes’ and ‘sizes’, that the borrowers come in, could be ‘measured’ before they even apply?
There’s the conventional way of assessing ability and intent, that somebody goes and checks everything out. There is a chance that some intangible insights about the applicant’s neighbourhood get captured too. But there’s a catch — these insights are mostly qualitative, exist in silos, and are rarely utilized to generate meaningful datasets. Location insight is quite an overlooked scenario, conceding to the idea that most of who we are, the behaviour we operate with, and the influence we acquire are dependent on our location.
In fact, the attributes a location exhibits, such as transport network, availability of hospitals, schools, types of retail brands, rentals and more, determine the living standard of a neighbourhood. And location intelligence in itself is a self-correcting tool; individuals in any given term are weeded out, whether in business or in residence, by the value proposition a neighbourhood offers in contrast to the cost of ‘being’ there. Meaning, that individuals in high-income groups move out if they don’t find enough worth being in an area, and alternatively, others get pushed out if they can’t afford where they are at.
Figuring out how location impacts behaviour, here’s where most banks and NBFCs would end up. Firstly, if it could be assumed that ‘affluence’ — which is an aggregate of the intangible worth of being in an area — is correlated to loan defaults, could it be that affluence can determine how ‘risky’ it is for the borrowers to default on a loan? Secondly, could this ‘affluence’ be measured? And thirdly, is it communicable?
There are 1500+ data points that can be captured to score ‘affluence’ at a 100m x 100m level in any locality across India, 200x more granular than pincode level data. This data is available but is stored in large and complex datasets which exist in 800+ public repositories, and requires extensive mapping to make a system understand locations the way a human eye does.
This is why GeoIQ is a one-of-its-kind software service that has harnessed location intelligence to help Banks and NBFCs make real-time decisions on incoming borrowers. We at GeoIQ compress extensive data points to generate direct and communicable scores that predict a user’s affluence the moment their lat-long or ‘address’ is captured. What’s more, our machine learning systems consistently train using your datasets and can lead to immediate identification of the ‘riskiness’ a customer is coming in with. Could location intelligence help make lending practices better in India, given that 90% of us have no credit history? Maybe not fully, but it’s a great head start.
Our solutions help businesses reimagine what they’re up to, on something as simple as a map. To know more about what we do, or to reach out, check out www.geoiq.io.
GeoIQ Blog Newsletter
Join the newsletter to receive the latest updates in your inbox.