Alternative data the next frontier for lending tech
Written on the 30 October 2019 by Matt Ogg
"All the technology, processes and procedures that we build are constantly evolving, and I think that's what's separating us from the giants in our space and the legacy lenders," says Jacaranda Finance founder Daniel Wessels.
For most people the word association with "subprime" would be "mortgage crisis" - the cumulative bubble of too many uncreditworthy people getting access to easy money, the failure of the USA's biggest banks to assess risk properly, a proliferation of foreclosures and ultimately the GFC.
Jacaranda Finance founder Daniel Wessels was still in his teens as that financial carnage reaped destruction across global markets in 2006, and belongs to a generation that has largely eschewed credit cards while embracing neobanks and innovations like buy-now pay-later (BNPL).
Since 2014 Wessels has built a business on lending to individuals most banks deem "unbankable", paying close attention to compliance while harnessing tools that allow for a better prediction of creditworthiness.
"It's harder for us to lend someone three grand than it is to get a mortgage," says Wessels, who won this year's Brisbane Young Entrepreneur Award in the Finance category and will be one of many competing for the top prize at the national awards on 16 November.
"We have to go through a pretty regimented process by law - we have to look at 90 days of statements, we have to take into account income, all expenses, etcetera to get the full financial picture of where they're at, whereas banks are all chasing mortgages.
"Because they've [banks] got such good security backing the asset, they were happy to take risks, and they were doing cowboy stuff for a long time. But this Royal Commission really shook things up."
To make Jacaranda Finance's rigorous process feasible, the group has used machine learning and advanced algorithms "to be able to make those decisions much quicker than 99 per cent of competitors in our space".
"The second phase was getting that instant banking into play, and the feedback has been phenomenal - people are loving it.
"What's really important is there's responsible lending and responsible borrowing, so there's an onus on both sides of the fence. Obviously the 'unbankable' are considered vulnerable lenders."
But how do these algorithms work? Wessels summarises them as pointing the lender in the right direction, using bureau data like a bank would but also application data, and day-to-day transactional data to determine the how, where and why of the borrower's expenditures.
"We use thousands of data points to determine if someone is likely to pay back their loan."
"Our back-end machine learning platform, has automated what a human would traditionally have done to gather average income, expenses, how many dependants they have, and we've built that in line with how ASIC (Australian Securities & Investments Commission) measures minimum affordability - we use the Henderson Poverty Index.
"It's about dotting the i's and crossing the t's to make sure that we are being responsible lenders. The industry itself has a terrible reputation for people not being responsible, companies not being responsible, but the good news is ASIC are slowly weeding them out."
Alternative data the next frontier of lending
It's one thing to assess a loan application based on income or expenditure, but what about someone's behavioural patterns while they're actually filling out their online forms? For Wessels, this question is part of a broad-reaching approach to using data in order to make better decisions more quickly.
"Keep in mind consumers are quite savvy. They might have applied somewhere else and were declined because they were overcommitted," he says.
"We might ask someone how many dependants they have and they put in four, and then they go to the next step and back and change it to two. Then they might change their weekly income as well.
"We're using identifiers and behavioural patterns in the application process, and behavioural patterns in the bank transactions."
To give a hypothetical scenario, he says over time big data might conclude a combination of a certain number of amendments in your application process, the time it takes you to fill out the online forms, the phone you use and even where you eat might make you more of a risk.
"It's picking those behavioural patterns and picking the negatives and positives out of them. And the way we do that, it's all about data capture," he says.
"Every time someone interacts with us, we store that somewhere - that's where it's leading.
Jacaranda Finance also provides lead generation for other lenders, and Wessels says some of them use more manual methods like checking social media to check whether what an applicant says is true. But he is also looking to developments in other markets - partly in the US but certainly in third-world countries where everyone has a smartphone but there are no traditional credit bureaus.
"What they're doing is using social data to score people to the point where Jim Smith is friends with Bob Jane down the road, and we know that he was terrible at paying back loans; they're links that are affecting your ability to borrow money.
"It's guilt by association but it doesn't stop there. Do you know who is looking at your browser history and where are you shopping?
"We've got a little operation in South Africa where we do some lead gen (generation) in this space because the bureaus don't hold enough data on people. There are two big indicators at play - there's the capacity to repay a loan, and then there's the question 'do they actually want to repay the loan'?"
In the future, he expects loans will be made based on big data assessments of character to reach an approximation on that second question.
"In South Africa the biggest indicator whether someone has the character to repay is mobile phone records, and if they are paying their mobile phone bills on time.
"Using alternative data to make better decisions is the name of the game."
Business News Australia
Author: Matt Ogg