PERSON OF THE WEEK: As federal regulators and the nation’s largest secondary market investors continue to seek out ways to make home financing available to more Americans, the traditional methods of evaluating consumer credit are being called into question.
Increasingly, lenders are seeking out alternative methods to assess the risk involved in lending to prospective borrowers with non-traditional credit scores. Long dominated by a handful of consumer credit repositories and one primary competitor, Fair Isaac and Co. (developer of the FICO score), the industry is finally evolving.
Maria A. Hatfield, Ph.D., is managing partner of Dattell, a big-data consulting firm. Mortgage Orb spoke to her about why she is bringing a new firm into this business and what she feels her new company can offer the banking and lending industry and the consumers it serves.
Q: What’s this business all about?
Hatfield: Any text-based data can be sorted, indexed and visualized to make it searchable and graphable in real-time. This enables lenders to track any aspect of their business either as an individual entity or in aggregate with other formerly disparate data sets. For instance, lenders can review how many loans resulted from an online marketing campaign and what the default rate was on those series of loans, informing future marketing efforts.
From an underwriting perspective, I see big data benefiting both the borrowers and the lenders/investors. It benefits borrowers because this could give them a new way to prove their ability to repay loans outside of their credit scores. For example, they could offer up access to personal accounts that demonstrate their ability to repay or give context to the less favorable portions of their credit history. Big-data integration systems would allow for the efficient import and visualization of this kind of information.
The lenders/investors benefit because they can learn more about a potential borrower before giving him/her a loan. For instance, we can track activity to determine if a person is already considering refinancing options, resolve questions of spending behavior and look for any red flags in a person’s online history, e.g. membership in hate groups or other organizations that the institution would not want to be associated with.
Q: Where did you get the idea?
Hatfield: Many other industries are capitalizing on this information for marketing purposes and customizing their products to consumer behavior. My partner and I witnessed this firsthand. We saw the opportunity for using this same kind of information for evaluating borrowers. If marketers use online consumer history to sell cars, why shouldn’t lenders use that same information to evaluate consumers for car loans?
Q: What do you hope to accomplish in this industry?
Hatfield: The lending industry has a reputation for not keeping up with the latest technology, but it doesn’t need to be that way. I hope to shake up the status quo and help these institutions with managing their internal data and their customers’ data in a safe, efficient and beneficial way.
This new technology could help increase access to loans for borrowers with non-traditional credit histories, while at the same time allowing lenders to learn more about borrowers. It is a win-win for both parties.
Q: Why do you think lenders should be looking at the data so deeply?
Hatfield: To start, lenders should be interested in big data integration for the same reason grocery chains, universities and tech companies are – to improve performance. This kind of reporting is used to quickly review all aspects of a healthy company, including customer experience, retention and attrition; employee behavior; monitoring and resolving technology errors; reviewing marketing campaigns; following sales; and much more.
Specific to the lending industry, this technology can improve the way we underwrite loans. Lenders that hold their loans in portfolio should be interested because we are providing them a way to learn more about borrowers, potentially reducing runoff and default.
But most lenders sell their loans, and in those cases, the interest lies in the secondary market. Investors purchasing loans should be more demanding about the information they receive from lenders when they buy loans, and we’re already seeing that trend. Why would any investor settle for a handful of numbers when there is more information available than ever before and increasing daily?
Q: What exactly is “big data” as far as mortgage lenders are concerned? How do you define the term? What types/varieties of big data might lenders be using in the near future in order to underwrite loans?
Hatfield: Big data includes both information internal to a lender, and information that is public or associated with a specific borrower. Internal data is traditionally siloed within departments: marketing keeps their stats, sales their own, loan officers have a separate record system and IT another. With big-data integration, all of that information is automatically and securely transferred to a central location – either in the cloud or on private servers – and made searchable and graphable in a dashboard.
For instance, a lender could easily track the success of a marketing campaign by looking at how many loans were issued as a direct result of that effort, and a lender could also monitor how a technology error affected loan origination. And importantly, alerts can be designed to notify parties when certain thresholds are met, such as an increase or decrease in website traffic.
Big data that is located outside of the institution includes all publicly available information and all private information that a borrower shares, e.g. email and search history.
Q: What is the cost of integrating the big data in with a mortgage lender’s underwriting system? Are most lenders paying third parties for their big data (as a service) or are they using public sources? Can you give some specific examples of private third party big-data providers – a public source of data – that mortgage lenders might use today, or in the future, to underwrite loans?
Hatfield: The cost of big-data integration includes both the fees charged by experts to set up the infrastructure and that of running the servers, either internally or in the cloud. Consultants use open-source tools – the same tools used by Fair Isaac and other companies – to create custom solutions for each individual lender. This is the most economical and tailored way to build, fix and/or maintain a big data integration system.
There are third parties that offer proprietary software that charge expensive monthly fees that increase with the amount of data ingested. Additionally, these third parties frequently require institutions to share their data with them, creating a significant security risk. Those charges and security risks are eliminated with open source tools that enable lenders to run the system on their own infrastructure.
Q: Considering there is a cost involved in using big data to underwrite loans – and lenders are always struggling to keep the average cost per loan down – what would you estimate is the return on investment (ROI) for using big data as part of one’s underwriting program? What are the advantages of using it in terms of risk, QC and profitability?
Hatfield: Lenders are embracing big-data underwriting tools because it gives them access to borrowers that they cannot reach using traditional consumer information services or traditional underwriting efforts. Since big data makes more information more readily available to lenders, the costs are typically lower per loan than traditional methods. Since lenders can’t even get this business without it, we typically talk about the opportunity cost of not employing these tools rather than focus on the likely modest reduction in overall cost per closed loan that would result from adopting them.
Q: What are the implications of incorporating big data into underwriting, in terms of recruiting and staff training? Is it important that underwriters be properly trained on how to use big data?
Hatfield: There is very little training required to take value away from the dashboards that visualize the data. It would be similar to getting used to a dashboard in a new car: training in minutes, not hours.
On the other hand, the training necessary to build a system would be prohibitory for most institutions, and that is why they seek outside consultants to efficiently and cost-effectively build and maintain their systems for them.