For generations, banks held a near monopoly on lending money to consumers. The rise of the Internet opened many new possibilities, including the advent of peer-to-peer, or “marketplace” lending.
New apps, new procedures and new ways to use big data are boosting the capability of the latest group of non-bank lenders to provide loans to folks with slight, bad or no history of credit.
Banks are in the alternative lenders’ cross-hairs. They account for about 3 percent of the $1 trillion market for business and personal loans. Goldman Sachs estimates that non-traditional lenders may grab 7 percent of annual U.S. bank profits within five years, worth about $11 billion.
“Big data” is lingo used to describe computations that sift immense data sets for valuable nuggets of information. Google was a pioneer in the use of big data and artificial intelligence techniques when it developed its search algorithms. These algorithms collect and process hundreds of thousands of data points to arrive at the content and sequence of search results.
A group of enterprising companies realized that the same kind of processing could be used in the field of consumer credit by developing new underwriting methods that went beyond FICO scores.
Not all types of big data are equally accurate or valuable as predictors of creditworthiness. The Fair Credit Reporting Act has compliance rules for the big data used by employers, insurers and providers of credit, among others. Big data used by creditors also must comply with the Equal Credit Opportunity Act to protect vulnerable groups.
The newest development is the use big data from mobile phones to establish the creditworthiness of prospective borrowers by pulling and evaluating smartphone data via innovative algorithms. Individuals can receive modest loans merely by permitting loan providers to analyze their smartphone data and determine whether the consumer will likely default on the loan or pay it back.
Companies in tech hubs like Silicon Valley are using non-traditional sources of information to assess the credit reliability of many consumers. The latest group of phone-reading programs utilize this process in emerging markets, where a lot of people lack FICO scores or credit histories.
The National Consumer Law Center highlights some interesting alternative data points:
- Rental data: Does the data include positive activity, or is it limited to delinquencies? Does it incorrectly downgrade credit scores when tenants withhold rent due to disputes with landlords?
- Utilities data: While most utilities only report customers that are substantially delinquent, a move to include customers who are only 30 or 60 days late may have a negative impact out of scale with the offense.
- Subprime credit: Often, payday loans and other forms of subprime credit go unreported. A shift to reporting these might unduly disadvantage subprime borrowers and increase their vulnerability.
- Telecommunications data: Sourced from phone and cable companies that have fewer consumer protections than do utilities, the inclusion of this data might be misleading, especially if consumer disputes are involved.
American companies that make use of non-traditional methods to determine creditworthiness include ZestFinance, LendUp, Affirm and Lenddo.
ZestFinance’s underwriting system utilizes “big data,” or algorithmic analysis of huge quantities of data, to enhance loan decisions or furnish advice on people who have scant credit background.
Zestfinance.com algorithms were formulated by employing information specialists from Capital One and Google to find various factors that correlate to a consumer’s financial trustworthiness.
The company claims that its underwriting models improve results — that is, enable more accurate credit decisions — by 40 percent, thereby improving credit access and loan repayment rates. Dozens of separate models can run in parallel to quickly make underwriting decisions.
In addition to operating its own online lending site, Basix, ZestFinance markets its services to alternative lenders who want to improve decision quality, expand market penetration and maintain a low default rate. The company now boasts of more than a hundred “smart geeks” on staff exploring innovative ways to exploit big data.
LendUp is a direct lender that extends short-term credit to subprime borrowers using techniques that go beyond FICO-score underwriting. Its algorithms offer points to debtors who attend classes for credit education and who develop a good track record of repaying loans.
Annual percentage rates (APRs) begin at 29 percent. As you rack up points on the LendUp Ladder, your APR decreases and loan limit increases. Borrowers who achieve high point accumulations have their loan payments reported to the credit bureaus, an important factor in boosting credit scores.
Examples of events that can earn you points in LendUp’s system:
- 125 points for the initial loan
- 125 points for completing each free education course
- 500 points for referring new customers
- 1,000 points for repaying you loan on time, up to 1,000 points per month.
Affirm offers a modest line of credit to customers via a variety of non-traditional and traditional credit analysis methods. For example, Affirm requires access to your bank account to help determine ability to repay. The company targets millennial buyers, many of whom prefer to use debit cards and thus don’t have an extensive credit history.
Customers access the Affirm smartphone app when checking out at a store that accepts Affirm. The procedure requires the customer to enter some basic information, choose among Affirm’s repayment options and accept the loan. Approval decisions are made instantly, and APRs range from 10 percent to 30 percent. Affirm guarantees payments to merchants who sign up for the service.
Lenddo makes use of more than 12,000 data points gathered from social websites, such as Yahoo, Google, LinkedIn, Twitter and Facebook, to assess a consumer’s potential to pay off loans. It targets “thin-file” customers — folks with sparse credit histories.
Lenddo doesn’t make loans, but rather offers underwriting information to lenders. Its app is installed to front-end a website or mobile app, and can be used in Facebook. The company claims to have enabled more than half a million loans worldwide. Lenddo now has offices in New York, Manila, Bogota and Mexico City.
Kabbage offers small businesses credit lines from $2,000 to $100,000 using alternative data, such as business performance and bank account information. The company claims to process 95 percent of its loan requests without human intervention. Loans must be repaid in six months or less.
Kabbage claims to have funded more than $1 billion in business loans. Fees are 1 percent to 12 percent for the first two months and 1 percent for each of the remaining months, without prepayment penalties. Other small-business lenders using automated algorithms include On Deck and Lending Club.
Upstart is a peer-to-peer lender that augments FICO-score underwriting with data about an applicant’s job history and education: schools attended, area of study, academic performance, etc.
It uses algorithms that simulate more than 50,000 scenarios and analyzes the results to assign a risk rating and APR to an applicant. Upstart claims its APRs are 30 percent lower than those of conventional lenders, and that 98 percent of its loans are either paid in full or current. The company arranges loans of $3,000 to $35,000 with 3-year terms. Their lowest APR is 4.66 percent.
Upstart claims to have arranged more than $250 million in loans since it began operations in May 2014. The average FICO score of borrowers is 692, and average income is $109,645. About 91 percent of borrowers are college graduates and roughly three out of four use loan proceeds to refinance credit cards.
The newest apps by outfits such as Saida, Branch and InVenture are installed on a prospective borrower’s smart device, where they gather all kinds of information from emails, texts, GPS coordinates, retail receipts, social media posts and other sources. The apps evaluate many kinds of alternative data, such as:
- How frequently the device is re-charged (frequent recharging indicates the user is a responsible person)
- The number of miles the device owner travels each day (traveling to diverse locations supposedly indicates you are more dependable)
- How many text messages the owner receives (more texts correlate to better creditworthiness)
- Whether the contact list includes last names (putting in last names is a plus)
- Whether consumers hold off making phone calls until night rates are in effect (indicates a price-savvy consumer, a good thing)
- Whether or not consumers are gamblers (gambling is regarded as a plus in this circumstance)
- Whether the applicant uses all capital letters to fill out an application (that’s a no-no).
The new apps make use of these and a host of additional factors to forecast credit reliability. They share similar practices, such as quick decisions and transfers of money to customers’ mobile wallet accounts.
Customers repay in installments, with no prepayment penalties. Currently, these apps are employed mainly in emerging markets, including Nigeria, Kenya, South Africa and Tanzania.
Every one of these nations boasts significant mobile phone adoption, especially South Africa (34 percent) and Nigeria (27 percent). Up to a third of African cell phone owners read or speak English. The use of smartphones in Africa is skewed towards 18- to 34-year-olds.
Customers need to be prepared to allow the new applications the freedom to collect many data points from their smart devices so they can be used to formulate alternate credit scores. Big companies, such as IBM and Visa, have entered this field by developing their own proprietary apps.
Cautions and criticisms
The adoption of alternative data as a means to extend credit can help the 26 million Americans lacking a credit history and the 18 million who cannot be assigned FICO scores for lack of timely data. The challenge is to incorporate the alternative data where justified without expanding its use, and thereby possibly discrediting, consumers with adequate scorability.
Also, in some cases, no score is better than a low score. The latter might create problems for consumers, such as getting a job, and may open low-score consumers to targeting from predatory lenders.
Some of the questions and criticisms around the use of big data for loan underwriting include:
- The cautionary tale of Countrywide Financial, which employed big data extensively and collapsed during the 2008 financial crisis. It was taken over by Bank of America, which eventually paid $1.3 billion in fines due to improper loans.
- Big data may create complacency because it could be manipulated. The objectivity of the data depends on how a big data app is programmed.
- Big data algorithms are complex and are thus hard to regulate.
- Underwriting based on big data has been criticized as opaque and subject to incomplete or incorrect data.
- There are concerns about the potential for discrimination against blacks and other minority populations whose members have historically been more likely to have no, or low, credit scores.
- Will big data help reverse centuries of discrimination, or will big data perpetuate the socioeconomic divide? Can big data level the playing field, or will minorities continue to pay much more for financial services than do their white brothers and sisters?
- Are big data predictions fair? Do they account for extraordinary changes to a consumer’s life? Some think credit scoring lacks the finesse to judge creditworthiness correctly.
The rise of marketplace lending hasn’t escaped the notice of regulators, who are exploring how sturdy these companies will be in the inevitable economic downturn, and whether new rules are needed to protect investors and borrowers.
This summer the Treasury Department put out a request for public comments on the business models employed by marketplace lenders, expressing interest in the accuracy of their underwriting models, among other things.
Credible is a multi-lender marketplace where lenders including Avant, LendingClub, PAVE, Prosper and Upstart compete for your business. You can compare personalized offers from multiple lenders without sharing your personal information with lenders or affecting your credit score. Eric Bank writes about small business, personal finance and science. He holds a master’s of science in finance from DePaul University and an MBA from New York University.