In recent years Non Performing Assets have emerged as a major headache for banking organizations. According to a recent report, bad loans have touched a new high of 9.8 lakh crores. If the economic cycle doesn’t pick up fast enough then the amount of bad loans can increase exponentially in near future. Although large corporate houses account for the major chunk of NPAs but loan defaults by SMEs is also a cause of concern for financial institutions. NPAs accumulate because businesses are not able to generate the expected cash flow from their business and do not have any other resources to repay their loan. This can happen due to a variety of reasons like:
- Disruption in supply of raw materials
- Sudden decrease in global price of finished products
- Operational costs going out of control
- Loss of money because of fines imposed by regulators
- Bad economic environment
- Business maintenance issue
To recover the value of the loans, banks have to start the process of taking possession of a mortgaged property when the mortgagor fails to keep up their mortgage payments. This process is known as foreclosure. Foreclosures are painful for the banks. Suppose the mortgaged property is the land from where the business is operating then it is difficult for the bank to calculate how much it will cost to improve the structure or bring it up to habitable standards so as to sell the property and recover costs. As a result, it is important for banks to avoid foreclosures.If the banks can predict borrowers at risk of foreclosures in advance then they can apply appropriate remedial methods to avoid this money leaking process.
Predictive analytics can solve this problem. A Financial Service Company which offered an integrated suite of financial services like Loans to Small and Medium Enterprises and Housing Finance wanted to reduce LAP foreclosures it lent to SME sector through “Working Capital Loans” and “Loans against Property”. The client utilized the data available to build a model using machine learning to predict SMEs with high probability to foreclose LAPs i.e. predict loan attrition. The data used was from a variety of sources like:
The client collected the above-mentioned information of every customer for the past 1.5 years and removed the customers who were less than 180 days into the system. The client didn’t have enough data points available for these relatively new customers, as a result, they had to be excluded from the analysis. As there were a lot of data points available so the client had to identify the relative importance of all these points in relation to the foreclosure rate. So the variable importance was determined and the predictive model to gauge the probability to default for each customer was calculated taking into account the relative importance of each variable/data point.
Finally, the client was successfully able to identify SMEs at risk of foreclosures 1 month in advance to apply relevant remedial measures so as to reduce foreclosures.
The client also harnessed the power of the given data to find out which tier cities and agents had the highest number of incoming foreclosure requests. The client analyzed the effect of repo rate and Sharpe ratio on the foreclosure rate. It was observed there was an increase in actual foreclosure rate for mortgage when the repo rate was decreased and vice versa.
By now I’m sure you want to save your organization from the painful process of foreclosures.Don’t worry we will help you out. Write to us at email@example.com and we will figure out the right course of action for you.