In funding small businesses, data science is sometimes just common sense

His company has lent nearly $3 billion to small business owners in the past 14 years, but Capital Access Network CEO Glenn Goldman doesn’t really care about credit scores. Unlike some Silicon Valley-entrenched personal lending startups, he doesn’t really seem care too much about data science, either — at least not by name. To Goldman and his team, figuring out who’s a good candidate for a loan really comes to good business judgment presenting itself as data.

Data science by another name

Although Goldman doesn’t talk about what Capital Access does in terms of data science, that’s essentially what it does. The models traditional credit-rating agencies use are designed primarily for speed, meaning they don’t always paint an accurate picture of applicants and tend to exclude a large percentage of small businesses from access to capital. So Capital Access built its own models using its own variables.

It’s trying to solve a problem, and it’s getting creative with data in order to do it.

For example, one of the ideas behind the formation of Capital Access — an “A ha!” moment that came as one of the company’s founders was having a hard time securing a loan for her Gymboree franchise — was that credit card sales represented a predictable stream of income that a lender could use to ensure they get repaid. Now, the company has a product called AdvanceMe that simply takes a flat percentage of a customer’s credit card receipts every month as repayment. The repayment model ensures that Capital Access gets paid back while its customers aren’t facing default by having to write oversized repayment checks in months when revenue dips.

But that type of loan also a great to learn about the income patterns of various types of businesses. “It created an infrastructure for gathering a lot of rich information about that business …” Goldman explained. “We receive a small amount of payment each night an an enormous amount of information.”

In fact, he said, Capital Access recalculates its models every night based on the information it receives every day from credit card transactions and other data sources.

Poor financial planner, or savvy businessperson?

What it has discovered over the years is that there’s a lot to be gained from applying common-sense approach to what has traditionally been a contextless process. According to Goldman, there’s often a disconnect between what credit-rating companies such as Fair Isaac and Experian penalizes and what those things actually mean. If a small business owner calls four banks in a day trying to secure a loan for suddenly available property to help him expand, for example, or if he runs up credit card debt to buy up inventory on the cheap, those might actually be smart business decisions despite the negative effect on his credit score.

That’s why Capital Access looks well beyond the credit score, even for more-traditional loan offerings than AdvanceMe. Goldman said it will look at factors such as how long a business has been around, the type of business (e.g., ice cream shop), where it’s located (e.g., Miami versus Chicago) and the owner’s business experience. Recently, he noted, Capital Access funded a restaurant that a bank turned down because it was technically a “new business,” although the owner was a successful businessman who sold a successful restaurant franchise so he could open his own.

When Capital Access decided it wanted to offer an entirely digital option (that product, called CapTap, launched on Monday), it had to find a way around a part of its scoring model that relied on data derived from actually contacting an applicant’s landlord to ensure good standing. CapTap promises a 7-minute approval process, Goldman explained, but reaching a landlord might take days. So, it tested the risk associated with this variable by eliminating the landlord check for a subset of customers during a fixed period of time. Write-off rates turned out to be quite a bit higher among that pool of customers, so Capital Access adjusted the CapTap model and origination fees in order to account for the increased risk.

Getting started with CapTap
Getting started with CapTap

Success breeds attention from large banks

The proof that these alternative risk-management models actually work shows up not just in lower default rates, but also in the attention they’re increasingly getting from large banks. Capital Access, Goldman said, actually has partnerships in place with banks such as Wells Fargo in which banks refer applicants to Capital Access or, if the credit score is within an acceptable range (but still lower than the typical 700 minimum required to get traditional financing), use Capital Access’s model to make their own alternative loans. Apart from the added interest revenue that new loans bring in, he said many small businesses use large banks for other financial products (e.g., checking accounts or credit cards), and those banks don’t want to lose that business by declining loan applications.

Goldman tells the story of sitting across the table from a huge international bank and explaining how Capital Access had lent $234 million to that bank’s customers without the bank seeing a single cent in commission fees or an iota of goodwill from helping someone secure a loan. One of the bank executives was shocked by a slide showing an average credit score of 561 (“We’d never fund them!” he exclaimed) and then even more so by the next slide showing a loss rate of only 3.2 percent — less than the rate on the bank’s own small business portfolio. It’s now a partner, Goldman said.

The ZestFinance data model
The ZestFinance data model

When I spoke with ZestFinance co-founder and CEO Douglas Merrill last year, he told me a similar story. He said his company’s machine learning-based model — which looks at approximately 70,000 variables in order to assess credit risk — is 40 percent more accurate than traditional models and that large banks and credit card companies we’re already signing up as partners in order to provide capital to lower-credit individuals.

“[If I’m a bank],” Merrill explained, “I don’t have to find those customers, I just have to serve them better.”

Merrill, by the way, will be discussing the world of alternative lending via big data at our Structure: Data conference next month in New York. He’ll be joined by Robert Frohwein of Kabbage, which lends money to small businesses that accept PayPal (s ebay) or sell on web marketplaces such as Amazon (s amzn), eBay and Etsy, among others.

Of course, the elephant in room with alternative lending is interest rates — namely, how to take on riskier bets without charging usurious rates like payday lenders often do. Goldman said Capital Access is sometimes competitive with traditional banks (charging APRs somewhere in the low teens), but does go quite a bit higher (around 40 percent) as risk goes up. It might not be ideal, “but the alternative is to say no to [some smart, passionate people with compelling stories],” Goldman said. “And our worldview is they deserve capital.”

Feature image courtesy of Shutterstock user yanugkelid.