Psychometric questions that are traditionally used for HR interviews are now redesigned to assess financial savviness and integrity of the applicants.
Data analytics is an integral part of everything Clix Capital does, says Katerina Folkman, head of AI and Analytics at the digital lending startup founded by Pramod Bhasin in 2016. The firm offers financing solutions across corporate finance, equipment finance and leasing, SME finance, consumer finance and housing finance, using technology to make loans faster, simpler and more accessible. In this interview with Banasree Purkayastha, Folkman talks about how Clix uses alternate data and analytics in credit underwriting. Excerpts:
What are the data analytics and AI models that you use for vetting customers?
We have developed a suite of Advanced Statistical and Machine Learning models in-house. We have hosted them in internal proprietary decision engine DELPHI. This engine currently governs all decision-making on every new application coming to our consumer portfolio.
When we started Clix Capital we had several collaborations with Bureau partners, to develop initial statistical models using their industry data. As our portfolio grew, we developed our own highly predictive ML models in-house though we continue to work with various external partners but usually as sources of alternative data. For example, we are working with GeoSpoc, that uses e-commerce data to score each geographical locality in terms of risk and prosperity. It helped us to make more sharp decisions, especially on New To Credit (NTC) customers.
We are pushing into AI territory too. Our Recommender Engine for cross-sell is based on reinforced learning techniques. We are also collecting and analysing psychometric and behavioural variables, to plug them into decision-making. Video analytics is our next big focus. We can collect video snippets of the applicants during the application stage, to use emotional sentiment analysis to predict willingness and ability to pay. AI and data analytics are a part of every loan sanctioned at Clix.
What is the scale of the data that you are tracking as of now?
We capture and track all traditional structured data in our data lake. But our data needs to be augmented with additional behavioural and unstructured data, in order to sharpen the analysis. This is why we look to on-board alternative data vendors—GSPs and supply chain marketplaces for SME products, and telecom, Geo, digital footprint for consumer, etc. We recently met a drone startup which can share valuable aerial data to confirm some points about our small business applicants. We track more than 500 variables about our customers right now, this will grow to more than 1,000 in near future.
How are you using alternate data to vet customers?
At the moment of underwriting, we assess and predict customer’s future ability and willingness to pay, as well as financial capacity to service specific amount of loan. Traditional proxy for the willingness is prior credit history of the applicant. Stable employment or cash flow to bank statements is another traditional proxy, for ability to pay. But many great customers may lie outside of what is outlined by these proxies. NTC customers may have good intention to pay, but cannot prove it with credit history. Some highly employable youngsters choose the entrepreneurial route, which makes it difficult to prove their “stable employment”. Analytics-driven lenders like Clix look for alternative but still reliable proxies, to assess them.
Psychometric questions that are traditionally used for HR interviews are now redesigned to assess financial savviness and integrity of the applicants. We have collaborated with an Israeli company to integrate their highly predictive psychometric questionnaire with some of our digital products. We use Digital Insights Score from Experian, developed to track digital footprint and online payment behaviour of NTC segment.
How big is your data analytics team?
We employ around 20-people in our Data Science + Data Engineering = Advanced Analytics team. Most of them are experienced data scientists, trained in Python and R. We will continue to hire more people as we grow our business.