Alternative data helps lenders score previously difficult-to-serve groups like thin- and no-file customers. Lenders seeking to serve those client groups need the right technology in place, Provenir’s executive vice president for North America Kathy Stares said. A provider of data and AI-powered risk decisioning software, Provenir serves companies around the world.
Provenir builds an orchestration layer that ingests data to help underserved groups be more accurately assessed. That’s the beginning of helping them generate a footprint and acquire products.
Leveraging alternative data
Alternative data is defined as non-credit-bureau data. Stares said it includes social data; individual financial data, such as cash flows from Airbnb; and socioeconomic data.
Provenir onboards this information, along with bank account data. They can input it into any decisioning or CRM technology. Together, they provide a good assessment and allow for the progression to lending decisions.
The trick is building technology that can productively process an ever-growing data supply. Machine learning helps banks and fintechs make better decisions.
“Taking all of the information and answering the question, is a new-to-credit or credit-invisible customer more risky from an onboarding perspective than the average consumer with trade lines?” Stares asked. “I would argue they’re not.”
The different types of alternative data
Stares said there needs to be more consensus on which data sources are the best risk predictors. She sees credit bureau data as overrated but is hopeful as bureaus are beginning to share real-time changes.
Information like job inquiries can improve credit risk assessments for thin-file consumers. Daily transaction data shows how folks manage their cash flow. Stares said this information strongly predicts how a borrower will handle credit instruments. That bodes well for the looming era of open data in North America.
Social data is also extremely valuable. How does an applicant interact on a social media platform? Who are their friends? What do they search for?
Dig further down, and you find fraud data. Is there consortium data? Are they applying for multiple accounts? Those are also risk indicators.
Technology supporting alternative data
Stares said that from a decisioning perspective, open data is the same as any other source. A good platform takes it and translates it into a sensible form like it would for any data set.
Machine learning or AI helps assess behavioral model validity, Stares explained. Run behavioral models through machine learning to see which ones are more predictive. That will inform how you should treat specific applicants.
“And what’s important is that whatever platform you choose, it’s native to the platform,” she advised. “That’s not going out to a separate entity. Having it native to the platform means that all the data that you put in is used to inform the model. That’s super important.”
Additional data can support thin-file applicants. Take bank statements and social data, feed them into a model, and see how an applicant compares against others like them.
“Cell phone payment is very predictive of credit behavior,” Stares said. “Do they pay online? Do they have a prepaid? Is their SIM card swapped from a fraud perspective? Have they been with the same provider for X amount of years?”
Stares said alternative data access becomes even more crucial when assessing folks with no credit history. Lenders can look at an applicant’s international footprint.
With some credit bureaus lacking international capability, social and lifestyle data can help. What bills are being paid from an account? Are there regular incoming payments?
Issues needing attention
Don’t dismiss any one data point, either. Stares said they could enrich decisioning models.
One challenge facing the industry is that alternative data effectiveness cannot be easily compared to key performance indicators, unlike traditional data. There is no assured correlation, only ones that “seem” effective.
“That causes concern inside of organizations because they want to say yes or no,” Stares said. “They don’t want to say maybe. That’s the challenge.”
Preparing to work with alternative data
How can companies best leverage insights from alternative data? How can they maximize their benefits from technology?
Stares said to be flexible and reactive to trends and macroeconomic demands. Consider your desired outcomes. Is it onboarding? Serving existing accounts? Detecting early distress?
There are best practices, and Stares said Provenir’s technology enables customers to respond in real time to such factors. Look to COVID-19, which forced everyone to make fast adjustments—those who did survived. Many fintechs didn’t.
Platforms like Provenir’s, buoyed by machine learning, allow financial institutions to pivot quickly.
“Our platform often sits on top of legacy software to enable you to respond and change your strategy, to challenge your strategy, to use ML capabilities that you may not have had before to respond to real-time events,” Stares said. “Today, there’s a macroeconomic slowdown. Nobody expected a bank to collapse.”