Mitigating Credit Risk Of a Flight Traveler

Introduction

Remember those childhood days when we all used to stare at an airplane flying across the sky and wondered how this world looks like when seen from the airplane. As we grew up, some of us could afford to fulfill that dream. However, for many of us, it’s been a dream so far. But maybe not anymore. Thanks to Mihuru, a travel finance platform that makes your flying affordable with its single-click credit. No funds in your bank account and you are not one of those fortunate people who have a credit card. No worries! You can still book your flight on credit. Yes! you heard it right, it’s for everyone. Freedom to travel in its truest sense. So, how is it possible? We at Mihuru made this possible with a combination of state-of-the-art technology, data science, and in-depth domain expertise of travel business. A single click credit is just the tip of an iceberg. The interesting part is what goes behind the scenes. A credit decision engine that approves a credit limit keeping in mind risk mitigation. Let’s find out what exactly happens there.

Defining credit risk for a flight traveler

In the case of a flight traveler, knowing not just about him but also about his travel habits is very critical to define the credit risk of a person. A person travels for various reasons. Enjoying vacations, attending special events like marriage, participating in festivals or sports events, and business trips top the list. Travel expenditure helps us understand the user travel index. Our proprietary algorithms make extensive use of user’s digital footprint, mobile device-based data, and third-party travel data aggregators to calculate user travel index. The algorithm is completely based on advanced machine learning techniques that learn on its own by churning more than 500 data points revolving around user’s travel habits. To list some of the important data points are average spend on travel, frequency of travel, a preferred mode of travel, purpose of travel, plus many more.

And, when you combine these data points with traditional information like your credit bureau history and your banking transactions, it sheds light on some of the important aspects of user credit behavior. Would you offer a travel loan to a person who has previously booked a flight for an exotic vacations in Hawaii while defaulting on his current credit card bills or a new to credit millennial who is frequently spending exorbitant amounts on lavish hotel stays while traveling outstation but unable to maintain a minimum required balance in his bank account and missing his electricity bill dues? Most likely not. Our algorithms try to find answers to questions revolving around the ability to pay and most importantly intention to pay, two most critical aspects in defining the credit risk of a person. The credit decision engine assigns a risk ranking to each credit applicant. Not to forget, all this action happens in real-time. So, how do we decide on credit approval, the quantum of credit, pricing of credit, etc? Let’s deep dive into the next section.

Credit risk mitigation techniques

The risk ranking is a reference number in a range (say 0 to 100) that quantifies the credit risk associated with a person. The acceptance or rejection of an applicant is a function of a trade-off between target approval rate and target portfolio delinquency rate. The accepted applicants are assigned with a credit limit which is dependent on risk score and current debt burden of a person. Someone who has scored say 80 out of 100 and has a current debt burden of about 20% is likely to get a higher credit limit than someone who has scored say 60 and has a much higher debt burden of about 45%. Again, the rate of interest for approved credit is defined based on the risk ranking of a person. Better the ranking, lower is the rate of interest. Apart from the pricing of credit, we also make smart use of the refund policy of the flights to mitigate credit risk and to control exposure at default. High-risk applicants are restricted to book only refundable flights with complete/majority repayment of the loan before the travel date.

Also, this entire decision engine that assigns credit limit, decides on the pricing and manages an overall portfolio credit health is monitored continuously. Any shifts in trends/patterns in any of the data point become a part of a self-learning process of the algorithm. The targets for approval rate and portfolio delinquency rate are dynamically adjusted based on the monitoring reports.

Role of machine learning and AI in risk mitigation

Have you ever wondered how is all this possible in real-time? Is only a human capable of mitigating a credit risk of flight travelers? The answer is No. With the progress of machine learning and artificial intelligence, now machines have already surpassed human limits in terms of speed and accuracy of decisions. Our credit decision engine takes leverage of some of the best classification algorithms like KNN, gradient descent, random forest, etc as well as regression modeling. We are also trying our hands on some of the advanced techniques like artificial neural networks to uncover more hidden patterns in the travel habits of a person. So, what did we achieve out of this? About 95% automation in the credit underwriting process with a turn-around time less than 3 minutes, approval rates above 70% and total portfolio delinquency below 2%. Sounds impressive, isn’t it?

What’s next?

We recently completed a successful pilot for mobile-platform only instant travel credit in partnership with one of the leading NBFCs from the Mumbai area. India is a growing economy where you cannot ignore the significance of tier II and tier III cities in GDP contribution. It is an enormous and untapped market ready to get online and travel on their own terms. And, that’s where Mihuru wants to be next. Right at the fingertips of these aspirational millennials from non-metro cities and fulfilling their travel dreams on just a click of a button.