
As Pearl Zhu said, “We are moving slowly into an era where alternate data is the starting point, not the end.”
As discussed in the previous two parts, the methodology to understand whether a consumer is in distress needs to be unique for the pandemic. Repayment history and income cannot be estimated with reliability, and neither are they stable indicators. This is especially true for the retail segment, which historically had less NPAs but may see a drastic increase in the coming months.
Algo360 proposes digital Early Warning Signals (EWS) for this purpose, which can be periodically assigned to the current portfolio—it can be monitored automatically within a comprehensive system. Systems can be upgraded if such signals are incorporated into the INDAS109 staging decision process. This will allow for control of probability of default and minimizing of Loss Given default.
In Algo360, credit scoring based on alternative data is already used for approval and rejection of applications. Consumers are classified into different sets according to their risk (based on credit scores), and subsequently their digital loan application is approved or rejected.
EWS solution, built on top of Algo360, will help financial institutions make improved and informed decisions about their loan portfolio. This solution has a built-in feature that will trigger whenever there is a possibility of an account becoming delinquent once they miss a due date. It will, therefore, limit the chances of borrower default. These include:
EWS variables represent consumer behaviour after the loan application gets approved and behaviour is influencing the book. It has to allow to reduce efforts, i.e., fully on the truly risky assets. Algo360 proposes the following approach, which would also loan underwriting process for digital lenders as well as borrowers. POC of unsecured portfolio on X-bucket portfolio EWS score divides consumers in three different segments depending on their risk. From loan disbursement to loan repayment, the most crucial task is to chase consumers. Without using EWS, 100 percent of consumers will be eligible for follow-ups indiscriminately (all those in the x bucket after missing a deadline). Using EWS, 30 percent of consumers are classified in the red zone, i.e., most likely to default on the current loan. So, we should apply significantly more efforts on that 30 percent consumer segment. For example, on a particular unsecured low-ticket size base, it was seen to be as high as 75 percent probability of default. Amber zone consumers are also in risk to default but not more than red. Hence, asset exposure will need to be reduced for this zone as well. If there are signs of over concentration in terms of a certain demographic group, over exposure needs to be further reduced. Around 40 percent of consumers in green zone—these consumers are more likely to return the loan. Also, green zone consumers can be used for cross-selling, thus helping in further loan disbursement. Usage of Algo360 EWS solution can significantly reduce time and effort on the riskiest lot as alternative data makes the discrimination between good and bad behaviour much more predictive.