Data Science & its Applications in Credit Risk

At slice, we are solving myriad of interesting and challenging data science problems on various fronts like Credit Risk, Fraud Sciences, Marketing and Product. Some of these problems are really core to our business especially credit risk and fraud and hence have been our key focus areas ever since we started our operations in Bangalore. In this article we would get deeper into what role data science plays in credit risk in general and specially at slice viz. at the time of credit line application from a user, evaluation and approval of the application, credit monitoring, collections, portfolio performance, portfolio expansion etc

Data Science at a startup is probably best thought of as more advanced RnD combined with rapid iteration and prototyping so as to ship a validated working model to production and can never be limited to just deriving insights through reports. Additionally, building a data driven engineering and operational culture is a major requirement for any data science team to thrive. Engineering teams need to get greedy about the vast amount of data that can be collected as early as possible. Storage is very cheap these days so the ROI from historical data is uncomparable when it comes to creating value out of it in future. For example at slice, we had started prototyping credit risk models even before we had launched and were able to figure out the kind of data sources that are relevant and hence need to be integrated in the product and data pipeline. Some of these data points only ended up in the form of production ready code 8-9 months down the line. We’ve been very fortunate to have a team that recognizes the value of data science without much convincing, so that has made the adoption of processes being built around our models and application of insights surfaced from the data science team very much seamless and agile. Producing strong results on both fronts has helped us further spread the data science love through the rest of the organization.

Being a very early stage startup in the credit payment space we decided to be over conservative on the risk side when we initially started. Considering the fact that expecting an income source proof from our target demographic could be a challenge in user experience and feasibility (which is a traditional Indian Bank’s cornerstone for evaluating credit risk) it was time we got a little more intelligent, specially when it comes to making use of all the relevant alternate data sources that are out there as part of credit risk modelling. For a legacy lending institution, the cost of sourcing, analysing, approving and monitoring a credit line which enables micro-credit payments at point of sale is not considered worth the ROI. slice uses machine learning and artificial intelligence on big data to its fullest extent to disrupt this formula thereby gaining a unique advantage over traditional lenders and hence is able to provide access to credit for college students and young professionals. This demographic constitutes of over 50% of India’s population and are the most technology savvy generation in the country. Being technology savvy and regular mobile users ensures we have enough data to play around with. Relying on our advanced risk models we are able to experiment steadily in the direction of portfolio expansion by choosing the least riskiest of the applicants from the chunk, which in the absence of a AI/ML based algorithms are bound to get missed.

Anything that includes human behavior, risk assessment and lending is extremely complicated but with the recipe that we are building at slice to solve this, its proving out to be very helpful in re-designing how credit is served to college students and millenials. We believe the best way to retain customers is by giving them an awesome experience over and over again. Instant gratification is something that millenials always seek as part of their lifestyle and gives them eternal happiness in this fast moving world. When it comes to evaluating credit risk we take special care in keeping our turn around times for loan decisions under control(Read hours and not days) and are always looking out for more ways to delight customers through fastest response time on their credit related queries. We constantly strive to build our product and internal approval processes around this thesis and the data science team plays a crucial role in providing production usable tools and visibility of metrics to individual domain teams hence making sure tight SLAs are met.

Another important business problem that we are solving through big data and machine learning is that of reducing costs in the approval process. Frugality is one of the core values at slice and data science team is no exception in absorbing this value and in driving us in that direction, by constantly coming up with insights and experiments to reduce the customer on-boarding cost. As a long term vision we feel credit assessment can be a purely online play and through successes from large number of tests being executed by the team in real world with real customers solidifies the fact that we are not far away from this kind of reality. Having less and less physical touchpoints ensures not just instant gratification to the users but also makes sure that costs involved in a decision are contained to the least.

In our opinion, credit monitoring and collections are as important in credit risk as credit approval is. Most Companies are almost always at a disadvantage when it comes to internal communication. The policies and processes that one team uses don’t always make sense for others, so staying on the same page can be hard with everything else going on throughout the day. Having an internal credit scorecard of each customer in place guarantees consistency of process within the credit risk, collections, operations and finance departments. By taking the same steps in each transactional situation, credit managers, risk analysts and their counterparts are able to operate seamlessly within the same guidelines. Developing an in-house credit scorecard backed by big data, machine learning and statistical analytics is one of the key projects being undertaken by the DS team at slice. The complexity here lies in the fact that each applicant or customer might have varied behaviors as well as needs of credit once he has spent some time in the system. Basing decisions on credit scorecard ensures specific needs of individual customers are met and outliers are detected with least Turn Around Times. Furthermore, it also gives a greater visibility and command over overall trend of loan portfolio and customer behavior once a customer enters and becomes part of the slice ecosystem.

Hope this article gives a brief overview of how crucial data science is at slice. We will be actively sharing more things that we are building or are happening inside through the medium of this blog so keep watching this space for more!

P.S. : If this seems exciting to you and you are interested in being part of the team, do checkout our current openings.

5 thoughts on “Data Science & its Applications in Credit Risk”

  1. whoah this blog is fantastic i really like studying your articles.
    Stay up tthe great work! You already know, a lot of individuals are looking around for this info, you can aid thnem greatly.

Leave a Reply

Your email address will not be published. Required fields are marked *