How Big Data Can Transform Consumer Finance
To invoke the memory of a familiar 80’s E.F. Hutton ad, when Douglas Merrill of ZestFinance speaks of finance, we should listen. (You might want to listen merely because he was previously CIO and VP of engineering at Google.) Midway through 2013, the Los Angeles-based startup received Series C financing to the tune of $20 million.
Zest for Lending – And Collecting
Merrill’s basic idea: use Big Data and machine learning to improve underwriting and collections in consumer finance. In other words, go beyond what he says are the 10-15 data points traditionally used for decisions like creditworthiness in a consumer’s FICO score.
The effort to get into the Big Data game for financial institutions is modest. They can load consumer data from many sources using Syncsort into Hadoop – the premier Big Data repository.
Some of the inferences Merrill makes from Big Data sound as though they violate the “correlation is not causation” maxim. For example, Big Data tells ZestFinance that creditors are more likely to collect on delinquent student loans if the borrower has comparatively more addresses after graduation – unless they move super-frequently. Similarly, borrowers who move far away from college are somewhat less likely to repay delinquent loans.
Merrill says this additional data helps collections outfits decide which loans are most likely to be repaid. The belief is that with a population chosen using these techniques, it’s more likely that lender and borrower can work out repayment plans.
In other words, these predictors may “just work.”
Cross Selling, Up Selling & Short-Selling
Fraud detection is complex and does highlight strengths of Big Data processing, but it’s unclear whether increased use of Big Data in these areas will affect consumers in any obvious way. In theory, outfits like ZestFinance will access multiple data sources for credit decisions, rather than using simple cutoffs, blacklisting and low-lend zoning of customer neighborhoods.
Cross-selling and up-selling can be made more effective by exploiting consumer purchase history or other data sources. Big Data could uncover illegal short-selling by exposing trading patterns. These “new” data sources may have been of interest to previous analysts, but were set aside as unfeasible.
Big Data creates an infrastructure that puts many options back on the table.
Whether consumers will respond to merchant messages from financial institutions is not completely clear. Banks already partner with merchants in credit card reward programs. Big Data could go far beyond this.
Banks could choose to favor car loans with specific car dealerships by creating loan application forms that are fully integrated with. Financial organizations, once thought of as operating at arm’s length from merchants, might regularly find themselves in a new position. Armed with Big Data and analytics – horsepower that dwarfs what a typical SMB could manage — they could participate in a wider variety of transactions than ever before.
Way back in 2012, the Economist reported that “Visa has teamed up with the Gap . . . to send discount offers to cardholders who swipe their cards near Gap’s stores.” The same reporter worried about incidents like the baby-care coupons mailed by Target before the teen had told her parents that she was pregnant.
These are not outlier examples.
Threats of a different sort could come from other sectors. Financial institutions could face disruptors like Zopa, a UK-based peer-to-peer lender that leverages Big Data to help both lenders and borrowers – a scenario fundamentally different from the one many financial organizations have with customers. So far, Zopa says it has helped consumers lend – and borrow — $806M.
A problem for solutions like this is that they could face new regulatory scrutiny. Problems with data quality will not go away overnight. Big Data-supported decision processes could create de facto discriminatory practices. For instance, it’s plausible that a police stop-and-frisk program, based on this or that weakly curated Big Data sources, could “recommend” profiling that might not cut legal muster.
In a 2013 talk at the Aspen Forum, FTC Commissioner Edith Ramirez gave voice to some of the concerns over Big Data:
The fact that big data may be transformative does not mean that the challenges it poses are, as some claim, novel or beyond the ability of our legal institutions to respond.
Having Its Way with Your 401(k)
Ideally, investors likely assume that they will benefit from Big Data analytics performed by their mutual funds. But some, like MarketWatch’s Chuck Jaffee, think more insidious trading data exploitation is at least possible. For instance, there is some risk that individual investors lose out when Big Data about their funds is given out to third parties.
Everything a fund does can and will be used against it by some sharpie armed with analytical tools and algorithmic forecasting abilities. It’s all perfectly legal, just shady; and because there is no way to quantify what fund investors are losing to the process, the entire practice is going mostly ignored . . . In short, if you can tell where managers, on the aggregate, are going, you can ride the crest of the waves they create. You don’t necessarily skin any one fund, but in a zero-sum game where a trader can guess where a security is going, they get their edge, and everyone who follows them into a security pays a higher price than they would have if Big Data had not lighted [sic] the path.
Big Data means new opportunities for firms. Consumer impact is less clear.
Consumer finance has been known for rewarding winners and punishing losers. At the time this story is being written, the average APR for fixed-rate consumer credit cards is 13%. The Fed Funds Rate is 0.25%.
Whether Big Data will move losers into the win column depends on more than who’s got the most Hadoop.