How long does it take to check 10,000 data points about a company? Digital lending firm FlexiLoans has built a machine learning engine for underwriting that can do this in 30 seconds.
"Considering that we sell loans without collateral within 48 hours, we need all the data points we can get about our potential customers," Abhishek Kothari, co-founder and head of product and technology at FlexiLoans, told TechCircle. "The idea is to make the digital borrowing experience similar to buying products or food on platforms like Amazon or Swiggy in 30 minutes.”
Founded in 2016 by four finance industry executives -- Kothari, Manish Lunia, Ritesh Jain and Deepak Jain -- the Mumbai-based non-banking financial company (NBFC) has granted more than 10,000 loans to small and medium enterprises with a cumulative value of more than Rs 500 crore over the last three years. While the average loan size is around Rs 2.5 lakh, more than 90% of these loans are under Rs 10 lakh.
According to Kothari, conventional financial institutions follow a 200-step process before approving a loan -- most of which are manual. But with technology as its secret sauce, FlexiLoans has sped up the process and reduced the cost of trawling through data points ranging from mobile numbers, addresses, PAN numbers, bank statements, credit history, location history, and nature of the locality, among others.
Apart from offering loans on its website, FlexiLoans has partnered with e-commerce platforms and Point of Sale (PoS) firms such as Pine Labs and First Data to lend to merchants working with them. Such platforms have large amounts of data about sellers, ranging from the sales value and growth of sales, among others.
"For us, 70% of our cost is technology and data science spread across both our front-end as well as back-end," said Kothari, who has been toying with data since he started his career as a business analyst with customer retention analytics platform Modelytics.
FlexiLoans has developed most of its technology in-house. It has around 60 software engineers at the moment and Kothari plans to increase this number by 25% in the near term.
"We are developing IP and have applied for patents on these technologies. What is my identity if not the credit models or the algorithms? These are long-term value creators and our competitive advantage," said Kothari, who has previously worked with Fractal Analytics, Barclaycard and Dell.
For instance, Flexiloans' OCR (optical character recognition) technology for fraud detection has now achieved 99% accuracy within just six months of training for the algorithm.
FlexiLoans also has a bad-loan ratio of less than 3%, which Kothari said was an indication of the robust nature of its analytical and ML engine.
Increased automation has also helped the company to disburse 15 loans per employee today from five loans per employee just three months ago.
"The data-driven decision model has helped us make many of the processes for a loan sanction now completely free of human intervention. Only when the algorithm raises a red flag does the human intervention take place or when required by regulation," said Kothari.
FlexiLoans has also created an algorithm to collect data signals for building a priority list of customers most keen on signing up for a loan. Kothari said that implementing this algorithm has tripled the volume of successful closures from the same number of customer calls in a day.
FlexiLoans is still deploying capital from its maiden fundraise three years ago. At Rs 100 crore, FlexiLoans had raised one of the largest-ever seed funding rounds. The startup’s investors include Sanjay Nayar, chief executive of alternative assets firm KKR India; Anil Jaggia, former chief information officer of HDFC Bank; and Vikram Sud, former head of operations and technology at Citibank.
Despite the NBFC sector facing a liquidity crunch, Kothari is hopeful that the overall ecosystem will help the company keep growing at a steady pace.
“We are not competing with anyone and have created a new market since around 50% of the loans have been disbursed to first-time borrowers. The long tail of the SME lending is the unserved or under-served," said Kothari, who pegs the overall MSME opportunity at $300 billion.