It's taken about three years of experimenting, but Thomas Bloetscher says Regions Bank has started to crack the code on lending to people with scant credit histories.

The Birmingham, Ala., bank has been sifting through reams of alternative data — mobile phone bills, rent payments, payday loans — to predict how so-called thin file consumers might handle a credit card or personal loan. The bank also draws on any information it might have about applicants' previous interactions with Regions or whether their household contains more than one Regions customer.

Bloetscher, who oversees the loan portfolio for the $118 billion-asset Regions, is well aware that other banks are parsing such sources for the same insights.

Across the industry, "we're all looking for ways to approve more people and grow households in the bank," he says. "You have to keep experimenting. The solutions we test today may be greatly improved in a year or two."

So what has Regions learned so far? For one thing, information that Regions can mine from its own systems outperforms the nontraditional data as a predictor of repayment. The realization has allowed Regions to approve applicants for credit cards whose FICO scores tend to be 20 points below what the bank otherwise requires.

"We've been able to layer in account information with consumers who have no credit and those tests have had much better results," Bloetscher says. "We feel there's a lot of power in those relationship factors. On their own, those alternative scores don't seem to help us much."

As the ongoing trials at Regions suggest, banks are beginning to look beyond consumers' credit files for information that may enable them to sharpen their predictive profiles and presage borrowers' propensity to repay financial obligations. Banks hope the information will help them identify creditworthy loan applicants and fine-tune pitches for credit cards, home loans and other products and services.

Though banks generally rely on as much information as possible to inform lending, they are learning what kind of payment data can help them when credit scores fall short. The discoveries are being fueled by insights from behavioral economics and data science, improvements in systems and software, and a slew of suppliers touting algorithms that allow lenders to more accurately assess someone's ability and inclination to repay.

[Conference Program Note: Join CFSI and American Banker for the 10th annual EMERGE Forum , with an agenda featuring the latest insights and strategies to help you meet the needs of underserved consumers responsibly and profitably.]

Banks stand to profit from the knowledge. As many as 65 million people — about 25% of American adults — lack a credit score or have a history that drags their score down, hikes the cost of credit or disqualifies them from borrowing, according to the National Consumer Reporting Association, a trade group. Payday and short-term loans used by consumers who have thin credit files or FICO scores that mark them as subprime generated roughly $84 billion in revenue from interest and fees, according to a survey published in December by the Center for Financial Services Innovation. And that estimate doesn't include home loans.

There's also an opportunity to bring more people into the financial mainstream. Many loan applicants, including immigrants and young borrowers, qualify for credit on more affordable terms when lenders turn to information not traditionally reported to credit bureaus. Research by LexisNexis has found that roughly 70% of consumers who lack a conventional credit score may qualify for mainstream banking products after factoring in nontraditional measures such as where they resided, degrees they pursued and real estate they owned (if any). About 20% of consumers qualify for credit on more favorable terms after adding payments for utilities to their credit files, according to a study published last year by Experian.

As the research reveals, conventional credit scoring incorrectly classifies some people who are otherwise willing and able to repay a loan. "You line up a thousand people and they got that score for a thousand reasons," says Zaydoon Munir, founder of RevolutionCredit, a company that amasses behavioral data about borrowers. "For people who are in the middle there's not enough differentiation."

A universe of information

Nontraditional information can round out the profile of an applicant, allow lenders to target offers more precisely or inform decisions in varied ways. Besides records of bills paid, such information may include patterns of purchases, measures of intent, inferences about character, schools attended, degrees earned, how you used your education, your social circle, your spelling and more.

The data encompasses a universe of activity and augments the information lenders can obtain from files housed by credit bureaus such as Experian, Equifax or TransUnion, which themselves gather information about outlays for rent, utilities, pay TV and telephone service. "We've always been interested in expanding the kind of data we collect on the credit report," says Brannan Johnson, who oversees data acquisition for Experian.

Though the forms of nontraditional data continue to change, the effort to gain insights into creditworthiness may be as old as lending.

"We found out 30 years ago if you buy life insurance you're probably a safer driver," Jamie Dimon, JPMorgan Chase's chairman and chief executive, observed in October.

Loan officers at Accion, a microlender in the Southwest, visit food trucks and landscaping businesses run by borrowers to size up their character, says Metta Smith, who directs lending for the organization, which uses conventional and alternative credit measures.

The use of nontraditional data has flourished in emerging economies, where companies such as Lenddo, First Access and Cignifi have developed platforms that match lenders' appetite for risk with information about people's lifestyles derived from factors such as mobile phone use, purchases of airtime and social media habits. "By 2030, 2 billion people who don't have a bank account today will be storing money and making mobile payments with their phones," write Bill and Melinda Gates in their foundation's latest annual letter.

Still, as information flows more freely and lenders centralize credit decisions, the methods of collecting nontraditional information have evolved. "The new community isn't necessarily a banker understanding everything that happens within 10 blocks of that branch," says Richard Crone, of Crone Consulting, a payments advisory firm.

The quality of nontraditional information continues to improve as data suppliers sharpen the algorithms they rely on to analyze people's behavior for clues to creditworthiness. Suppliers also are getting better at tracking payments by borrowers who rely on alternative lenders, according to Greg Rable, CEO of FactorTrust, whose company does just that to build credit scores.

According to Rable, that systematization has developed in the past five years. Previously the massing of nontraditional information "was brand new and there were so many points of data floating around," he says.

If anything, the imperative to draw on alternative data has intensified since the financial crisis, which triggered a pullback in risk-taking generally. Assets in U.S. banks rose by roughly $4 trillion since 2008 while loans have increased by about a quarter of that amount over the same period. "On the margin banks are failing to figure out how to lend effectively," says Patrick Reily, co-founder and CEO of Verde International. His company makes tools to model borrower behavior based on the relationship between someone's education, job, purchases, and other activities offline and on that mark the person as living a life Reily describes as "organized and structured."

Impediments to adoption

Despite the inclination among lenders to rely on nontraditional information, several factors have slowed its adoption. The Fair Credit Reporting Act and other laws that govern the investigation of creditworthiness filter the information that finds its way to credit files. Additionally, lenders who aim to sell their loans to a government mortgage entity continue to confront incentives to avoid making loans to borrowers who lack solid credit scores despite recent steps by federal regulators to ease the credit crunch for homebuyers.

According to some experts, banks also receive mixed signals from regulators who toggle between encouraging banks to draw on data that can expand lending and enforcing rules that discourage such activity. The Consumer Financial Protection Bureau's research office, for example, "is doing great work that's data driven" says Arjan Schutte, founder and managing partner at Core Innovation Capital, which invests in companies that create upward mobility. But the agency's enforcement arm can be "quite aggressive."

"The two are missing a corpus callosum," Schutte adds, referring to the part of the brain that ties together the left and right hemispheres. "In that kind of environment, banks that have had to double or triple their investment in compliance are appropriately nervous and reactive."

Technology, too, constrains the use of nontraditional data. Banks have to be able to extract information stored in varied systems, a task that can be a challenge, especially with information technology investments being pulled in varied directions. Suppliers of such information that lack distribution via a credit bureau also must find a way to deliver their data in a format that meets lenders' specifications while demonstrating compliance with a series of safeguards.

"Some of those companies are not as mature as an Equifax or a TransUnion or a regular reporting agency," says Bloetscher, who notes that, by law, consumers have a right to dispute the accuracy of information in their credit files and that credit bureaus, after being notified by consumers of such disputes, have an obligation to investigate. "Is there a dispute process? Does information get corrected?" asks Bloetscher, referring to suppliers of nontraditional data. "A lot of that infrastructure is not there."

Nontraditional data also must pass muster with risk managers, who look askance at borrowers lacking sufficiently detailed credit histories. In some cases their reservation may be warranted but for applicants at the margin the skepticism should yield to alternative data that demonstrates creditworthiness, experts say.

"If you know anything about how risk-averse banks are, you can imagine how challenging that conversation is," says Steve Ely, CEO of eCredable, which aims to help borrowers build credit histories from bills they're already paying.

Still, one legacy of the financial crisis may be that it prompted risk managers to get over their disinclination to give credit to the financially underserved.

"Lots of middle-American people, upstanding members of their communities, are struggling, and their FICO score has decreased and their ability to get access to mainstream credit has been challenged," says Schutte. "More risk managers today say, 'Let's figure out the goods from the bads.'"

Time plays a role too. As the work underway at Regions suggests, banks continue to experiment. "Adoption is process," says Ankush Tewari, senior director of market planning at LexisNexis. "It will take [banks] a long time to evaluate different sources."

Banks tight-lipped

Though banks may be approaching alternative data in varied ways, institutions tend to keep mum about their efforts. All but six of the nation's 19 biggest banks by assets either declined to comment or did not respond to inquiries on this topic. At the same time, 86% of lenders say they want to rely on more than the information contained in credit files, according to a survey published in October by VantageScore, which sells an alternative to FICO scores.

Competitive concerns may partly explain bankers' reticence. "Unlike traditional data, where it's very similar and highly overlapping regardless of what bureau, in the alternative data space you have much greater variety in data captured and covered," says Ethan Dornhelm, a data scientist at Fair Isaac, the maker of FICO scores. "It's a heavy lift to get to the point where you've identified the alternative data sources that are predictive and meet compliance standards and are robust."

Banks that discuss their plans publicly say that a mix of data informs their views. As part of a push to broaden its customer base, MUFG Union Bank has developed what it calls an economic opportunity mortgage geared toward younger homebuyers and those with low to moderate incomes.

As with most home loans, the bank examines applicants' credit scores, debt-to-income ratios, savings and other conventional measures of credit. For applicants who fall short by those measures, the bank also considers alternative sources of information, including utility bills, rent and insurance premiums paid. Last year the bank originated 961 such loans totaling $248 million.

"We can use...f different monthly obligations that won't report onto a credit report... and that demonstrate ability to make some sort of monthly payment," says Jonathan Bellomo, who manages residential underwriting for the $111 billion-asset bank.

Spokespeople for Bank of America, KeyBank and Ally Financial all said they consider a variety of factors, including traditional and nontraditional information from credit bureaus, as well as internal customer data, to make lending decisions.

Fifth Third Bank will lend medical residents who meet certain criteria the entire value of a home up to $500,000, an amount that increases to $650,000 for physicians who have been employed for at least a year. "This is an example of using alternative data to make an underwriting decision outside the traditional guidelines," says David Gunn, who heads mortgage lending for the $134 billion-asset Fifth Third.

Banks also are looking beyond their own labs and partnering with upstarts. Last May, MUFG Union Bank and Lending Club, which matches borrowers with lenders via the Internet, announced an alliance to develop products. As part of the pact, the bank buys bunches of unsecured loans from Lending Club that the latter underwrites based on its own analysis of creditworthiness. "Then, on a portfolio basis, we run our own models against that portfolio," says James Francis, the bank's head of consumer lending.

Nontraditional data also promises to allow banks to make changes beyond traditional underwriting and marketing. Verde, whose clients include JPMorgan Chase, Banco Popular and Royal Bank of Canada, draws on its models to help banks determine their allowance for loan and lease losses, as well as where to locate branches or which products to offer. Banks are turning to Lenddo, which infers creditworthiness from members' online social activity, to verify customers' identities.

Changes ahead for lending

In addition to the big banks, some community bankers are using nontraditional data to craft their own credit ratings. A slew of startups also are drawing on alternative measures of credit risk to challenge banks in lending to consumers and small businesses.

Whether the spur comes from startups or incumbents, lending looks likely to change as exchanges of information between borrowers and lenders speed up and lenders' ability to react to an applicant's assets, cash flow and other measures of credit accelerates.

Crone foresees a day when banks use nontraditional data to make loans that consumers tap via mobile wallets to purchase featured products at specific retailers. That might enable a consumer to purchase an appliance at Home Depot, for example, using credit that circumvents card networks entirely.

"The big data feed is allowing the lender to actually create a new loan type that's beneficial to consumers and to them," says Crone. "The issuance of transactional credit is a very appealing proposition."

Already lenders are using alternative data to fill in profiles of customers. The hope is to anticipate customers' needs at each stage of their lives. "That's a far different approach than 'let's make a credit card decision based on a three-digit credit score and move on,'" says Chris Atwood, vice president of product management at Equifax.

Jeff Stewart, Lenddo's co-founder and CEO, envisions companies specializing in pieces of businesses that today are the purview of the largest financial institutions. As he sees it, the activity will be propelled by a confluence of mobile devices, big data and a shift to transacting in real time that pulls in nontraditional information and slashes the cost of forging relationships with borrowers. "You will have companies that all they offer are savings accounts... or small business loans... and others that only help open accounts," he says. "Software driven, cloud-based services allow you to stitch together financial components much more easily."

Stewart imagines waiting for an elevator. "While I'm waiting, I should see what's going on with my finances, be educated and get updates," he says. "That 40 seconds, that's the future of financial services."