Deep Dive: How FinTechs Harness Machine Learning To Approve SMB Loans More Quickly And Safely

Small- to medium-sized (SMBs) are the backbone of the American economy: 30.7 million SMBs operate across the country and account for 64 percent of new jobs.

They represent 99.9 percent of all businesses in the U.S., but they face many hurdles. The U.S. Bureau of Labor Statistics estimates that 20 percent of SMBs fail in their first year and that approximately half do not survive for longer than five. Only 35 percent of such businesses make it past their first decade in operation.

The reasons for these failures are varied, but many involve cash flow issues, as 29 percent report they ran out of money and 18 percent cite pricing or supply cost problems. The ongoing pandemic has added new obstacles, too, as social distancing and stay-at-home orders cut off key revenue streams. Seventy percent of SMBs are adding new digital capabilities or enhancing existing ones to continue their operations, but firms often require financing to implement these innovations.

A multitude of SMB investors are looking to help, but lending to digitally focused SMBs comes with its own hurdles, including fraud risks and inefficient lending procedures driven by stiff regulatory measures. The following Deep Dive explores how these issues are hindering SMB financing as well as how FinTechs are exploring technologies to overcome these issues and give SMBs the help they need to thrive while the world economy faces its biggest challenge in years.

SMB Lending Risks

One of the most pressing issues plaguing SMB lenders is the risk of fraud. Bad actors deploy numerous techniques to scam lenders, and many ask for financing without any intention of repaying. A fraudster might steal a consumer’s identity and leave that individual carrying the bag or simply disappear after receiving a loan.

Fraud targeting SMB lenders has increased by 7.3 percent in the past 24 months, according to a recent survey by LexisNexis Risk Solutions. Small banks and credit unions (CUs) have been hit hard, experiencing losses of 4.5 percent of their revenues. Lenders are struggling to detect and stop these fraud attempts, with only 43 percent saying that they are very effective at identifying lending fraud. Forty-nine percent said that SMB lending fraud is more complex than personal fraud, but 59 percent devote the same amount of resources to both types.

Fraudsters are taking advantage of this reality as well as manipulating measures designed to ease legitimate borrowing like compressed lending cycles and multiple transaction channels to stage more fraud attempts than ever before. One technique known as “loan stacking” occurs when bad actors capitalize on digital lending practices by taking out numerous loans from different lenders in a short amount of time. This technique was impossible to attempt when loan applications required multiday processes that included visits to brick-and-mortar bank branches.

Another complication for SMB lenders comes in the form of government regulatory measures, which often handicap lenders’ abilities to simplify loans. The 2008 recession resulted in government regulators tightening the standards for creditworthiness, increasing the amount of checks that lenders must undergo to approve loans. These increased regulations make sense on the surface, as misguided loan practices were largely to blame for the recession, but they have resulted in a dramatic reduction in SMB loan originations. The U.S. Small Business Administration (SBA) estimated that SMB loan originations have dropped 40 percent from their prerecession levels, highlighting the crisis’ long-lasting effects on the business world.

Fraud and regulatory challenges will require lenders to double down on their efforts to safely and accurately gather and organize applicants’ data. A number of technological solutions exist to help them accomplish this, fortunately.

Mitigating Loan Approval Risks

Tackling SMB loan fraud — especially as more businesses go digital — requires dedicated efforts to spot and thwart cybercriminals’ schemes rather than relying on normal loan approval processes to detect fraud and halt suspicious applications. LexisNexis Risk Solutions’ study of SMB fraud found that 94 percent of businesses that were most effective at stopping such fraud had dedicated strategies to do so and used various technologies. Customer identity verification technologies like biometrics or two-factor authentication have been especially effective, forcing applicants to prove that they are who they say they are and reducing the number of fraudulent loan requests. Machine learning (ML) is another promising fraud-fighting tool, as it can detect slight discrepancies and inconsistencies in loan applications that human analysts might never notice.

ML is also effective at meeting government regulatory requirements and approving loans more quickly, which is a massive aid for businesses looking to quickly digitize. It can help lenders cross-reference applications and discover new information faster than humans ever could, and it automatically learns from these data sources to improve its predicative capabilities and generate accurate risk profiles for applicants.

Many FinTechs are using these ML applications to accelerate their financing approval processes. Each application contains a number of different variables that can impact the risk of the loan, including existing credit, past application histories and risk of fraud, with many traditional banks being forced to limit the amount of energy and attention that goes into applications. Not only does ML automate these processes and allow each application to achieve a much more thorough inspection in a fraction of the time, it also grows smarter with each application processed and applies this new intelligence to subsequent applications. This in turn leads to a much higher rate of loan approval that traditional banks can accomplish. FinTechs that harness ML to process loan applications were approving 57.2 percent of SMB loan applications during April 2019, for example, while 27.5 percent of these loan applications were approved by traditional banks that did not leverage the technology.

Even traditional banks are partnering with FinTechs in record numbers to access this technology, with J.P. Morgan Chase & Co., KeyBank and Scotiabank all entering into collaborations during the past few years. A report from PwC found that 82 percent of banks plan to follow suit in the near future, signaling a massive shift in the way financial institutions (FIs) harness technology to approve loans.

Reducing the amount of fraud and navigating government regulations will be essential to ensuring the success of SMBs for years to come, especially as the digital shifts that many of these firms have undertaken become permanent.