Debt collection is a very important part of the finance industry as it deals with collecting delinquent loans from the customers. Traditionally, the banks outsource loan recovery to third party debt collectors who do all the follow-ups on the delinquent customers to persuade them for loan repayment. It has been a proven model for the loan recovery process and is essential to minimize bank losses.

Away from traditional financing, there is a completely contrasting industry of microfinance that has evolved in parallel. Micro Financing deals with the lending of low ticket size loans to underserved customers who do not have access to traditional banks due to absence of a credit score. A 2020 report says that the global microfinance industry is expected to reach USD 196.4 Billion. These microloans are popular among the people in the developing countries of South Asia and Africa where many FinTech companies are operating to create the ecosystem of microlending as seamless and accessible as possible.

Microlending has its own challenges to take care of, the first challenge is how to make micro loans accessible to the unprivileged group and the FinTechs have adopted the model of making the digital loan available easily from mobile platforms via USSD/apps

But the next challenge faced by micro lenders is the debt collection part from these new to credit customers which is more tricky than the traditional debt collection!

Debt Collection Challenges in Digital Microlending 

1. Usually, the traditional loans are of high ticket value, hence it makes sense for the traditional banks to invest in third-party debt collectors for loan recovery. But the Digital micro loans are usually of small value; ticket size ranging from $10-300; hence hiring debt collectors is not feasible. So they have to rely on internal communication channels like phone calls, SMS, mails for loan recovery.

2. Most of the time banks act reactively when customers have started missing multiple repayment cycles. And by the time they initiate rigorous follow-ups customers often reach to a financial situation where they just cannot repay the loan and they start avoiding follow ups.

Due to the nature of loan, the digital micro loans have a huge volume. This also means there is always a high risk of defaults as well and this risk rises significantly in the current Covid-19 times where the businesses have slowed down and there are job losses happening everywhere. So the micro-lenders have to adopt new innovative measures for effective debt collection during the financial crisis of Covid-19. This is where machine learning and artificial intelligence can play a crucial part in augmenting the debt recovery process for the micro-lenders.

 

How Machine Learning can help in Debt Collection of Digital Microloans

 

 

1. Predicting the Delinquent Customers

The writing is always on the wall that the customer will soon default and the signs can start showing up early in their change in spending behaviors or the bank transaction patterns. Machine learning models are quite efficient in recognizing these hidden patterns. Hence banks can utilize the data of thousands of the previous delinquent customers and create a classification model to predict the probability that their current customers can also default. In fact, during this Covid-19 times, the chances of finding such high-risk customers are very high.

The customers with a very high probability of delinquency can be proactively reached out by the banks well in advance and start guiding them financially or give them financial options to mitigate the risk of loan default before it is too late. 

2. Alternative Credit Scoring

In the absence of traditional data, FinTechs apply data science and analytics on alternate sources of data like customer location, phone records, previous loan details, customers with similar demographics, etc to device alternative credit scoring systems for evaluating the repayment capacity of the customer. 

The alternate credit scores help microlenders keep a tab on high-risk customers with low credit scores and reach out to them proactively to guide them with better financial planning before they start missing their repayments. These customers are reached out at different stages of the customer lifecycle The alternative data credit scoring and data from the lifecycle journey of the customer is useful in creating the predictive model we discussed above.

3. Customer Segmentation

Not all delinquent customers suffer from the same financial crisis, some groups of customers might have certain common demographics or characteristics, others might have some other set of common factors. So it is very important to segment these delinquent customers and then implement debt collection strategies . For example, people who are old may require different follow-up tactics than young demographics; urban vs rural; salaried vs self-employed

Banks can apply unsupervised learning clustering techniques to create different segments of delinquent customers by using their demographics and banking data. Banks can then come up with different debt collection strategies for each of these segments. Such type of proper planning for debt collection would yield more fruitful results instead of having a general followup from the default customers.

4. Automated Communication channels

The financial crisis in Covid-19 is bound to see a huge number of customers missing their loan payments. Hence if there are multiple manual touchpoints in the debt collection process, it may turn out to be ineffective to meet a surge in defaults. Hence we automate the debt collection process as much as we can.

Microlenders can integrate their communication channels like automated phone calls, SMS campaigns, emails with the default loan prediction system. As soon as the system predicts the risk of default for a customer, the communications are triggered to give proper guidance.

Communication systems can also be made intelligent to contact the customers at the time of the day when they are more likely to respond. This ideal timing can be determined by analyzing their phone calling records.

Conclusion

Machine learning is not magic that can ensure successful debt collection, but it can surely assist banks in mitigating the risk in advance or help them in making appropriate loan recovery strategies with the right set of hidden insights about the delinquent customers. Since Covid-19 related financial crisis is going to stay for more time now, the role of machine learning in debt collection becomes more important than ever before.