FICO’s Sally Taylor-Shoff says there’s little evidence trended data helps lenders or credit card issuers
Credit scoring methods have evolved in recent years, but just because a formula is trendy doesn’t prove it’s effective.
Few are more qualified to weigh in on the latest credit assessment methods than FICO Vice President of Scores Sally Taylor-Shoff. And there’s much to talk about these days. In April, VantageScore announced it would begin using trended data – such as whether a cardholder tends to pay his balance in full or just the minimum each month – in its flagship scoring formula later this year.
Meanwhile, FICO has embraced unconventional scoring methods outside the U.S. It is now working with psychometric credit scoring pioneer EFL to help a Russian bank assess young consumers’ creditworthiness using online personality tests. Last October, FICO announced a partnership with scoring and identity technology firm Lenddo to help score consumers in India using social media data.
FICO also made a formal jump into alternative data with its FICO XD score in 2015. FICO XD uses consumer data from the National Consumer Telecom & Utilities Exchange (NCTUE) and public records to evaluate utility and cellphone bill payment information and other data on prospective borrowers.
But the firm that invented the most widely used credit score is less enthusiastic about trended data. One might surmise this is because FICO competitor VantageScore has gone all-in on it. However, Taylor-Shoff told us in a recent conversation that there’s not enough evidence that trended data helps lenders and card issuers on a large scale. Taylor-Shoff also provided an update on FICO XD and its global financial inclusion initiatives.
Q: What do you think about the inclusion of trended data in some mainstream credit scoring models? Do you think FICO’s traditional method could someday include trended data?
A: We’ve studied trended data, and we haven’t found that the return on investment (ROI) is there. There’s a lot of talk about being able to find some significant lift in small pockets. But for issuers and lenders, they need to see ROI across the board. They have to pay a premium for trended data.
When we say ROI, does the incremental lift – either the new customers you can get or how much you can save in loan defaults – justify the extra you pay for that premium trended data? The answer to that is no. The reality is there’s no differentiated information you get from trended data that you can’t get from looking at the snapshot of a consumer’s utilization.
Q: Can you comment on the overall adoption of FICO Score XD since it was introduced two years ago? What progress has been made?
A: Quite a few card issuers have been testing and validating the score, and we have the first early adopters already in operation. Credit is a space where issuers really have to validate and test something and then work everything through their compliance departments to make sure that they can accurately answer consumers’ questions about why they did or didn’t get credit. All of that is required by lenders, so it takes a while for them to adopt. But we’re very happy to see the early adoptions of the score in the market.
Q: Has FICO participated in the Consumer Financial Protection Bureau’s (CFPB) discussions about using alternative data to expand access to consumers who are “unscorable”?
A: We participated in their request for information. We focused a lot on what we learned with FICO XD and how important it is to have alternative data. There’s been a discussion about whether you can just lower the minimum scoring criteria to get at those (unscorables). Our research shows there’s just not enough data.
Some people are unscorable because they had problems in the past and then they stopped using credit. Their data is frozen at their time of distress. It would not be right to use that data without other alternatives like how they’ve been paying their regular bills. So, we did participate – we believe it’s important.
Q: Do you foresee any new developments in alternative data in the near term?
A: We continue to research other types of data. For example, we are looking at being able to better understand how a consumer manages their finances. We’re looking to other sources of data to get more cash flow information.
We’re also doing research outside the U.S. with data that’s even further down the hierarchy. Outside the U.S., where there are even larger populations that aren’t represented with some of the traditional data – more people who are completely unbanked – we’re partnering with other companies that are looking at things like psychometric data. People answer surveys that are based in psychology that can then predict how they view money and finances and so forth.
With another partner, we’re looking at someone’s social networks through their phone and their SMS. It’s not the content of what they’re saying in their SMS or their social media, but really the patterns of how they interact. We’re including that in a new score we’ll soon be launching in India.
Q: Do you think U.S. consumers will someday get credit scores using psychometric or social media data? Is the unscorable population in the U.S. too small (an estimated 53 million adults) compared to the rest of the world? Are there regulatory issues?
A: FICO XD as it stands now covers a little more than half of those unscorables, so we still have another 25 million to go. We continue to do research, and the NCTUE database is getting richer. We actually have done some research where we believe the next version will get to 70 percent.
Some of that lower hierarchy information starts to get into compliance issues. If we can get deep enough with what seems more reasonable to a consumer – how they pay their credit, how they do their finances – it’s much less likely that traditional lenders in the U.S. would tap that far.