Using Data to Speed Up Digital Lending: 6 Key Takeaways

In brief:

• Novel data sources like accounting, banking, and commerce data allow for faster, more granular underwriting by providing real-time financial insights.
• Introducing strategic friction through rules-based approvals and knockouts streamline pipelines to focus analysts on complex cases.
• Introducing new data sources in high-rejection areas allows low-risk experimentation without disrupting flows.
• Experts predict growth into new lending verticals and products as underwriting friction decreases.

‍

“When it comes to small business lending, speed to quote usually wins.” 

‍

Small businesses are growing more rapidly than ever. Last year, over 5 million new businesses were formed. These businesses have come to expect a more consumer-like experience when it comes to ramping up with vital financial products like credit cards, lines of credit, and debt financing. 

At the same time, these businesses are adopting technology-forward tools to manage their own businesses. Instead of using a bookkeeper, they use Quickbooks Online. Instead of a traditional bank, they use a neobank where they can apply online and open the account the same day. 

The rapid growth of small businesses has combined with those businesses' comfort with technology to open up a new frontier for banks, fintechs and other financial institutions. 

To find out how technology can allow businesses to win the all important “speed to quote” race, we recently gathered leaders from three key technology companies leading the charge when it comes to using data to underwrite small businesses financial products. 

Here were the findings. 

‍

The Panelists:

• Sri Srinivasan, Head of Risk at Ramp
• Prathik Naidu, Product Lead at Rutter
• Abby Grills, Product Manager, Middesk
• Jay Patel, Product Partnerships Lead, Middesk

New Data Sources = New Risk Strategies for Growth

Modern businesses are now more comfortable than ever using modern software and allowing financial institutions access to financial data such as their banking data, accounting data and sales and e-commerce data via APIs.

Now is the ideal time for financial products to innovate and make use of that data. 

The key to underwriting is determining your potential customer’s true financial risk. But now lenders are in a position to ask for not just documents and data from legacy data sources like Dun & Bradstreet and LexisNexis, but real time accounting, banking and sales data from the API-driven solutions that businesses use in their day-to-day operations.

The most common types of novel data sources lenders now have access to are:

• Commerce and Payments – You can see your potential customer’s revenue or sales data often in close-to-real-time and at very high granularity. If offering cash advances, revenue-based financing or inventory financing, a view into your customers’ commerce and payments data lets you view how their business runs down to the SKU level. The drawback of this type of data is that you don’t see your customer’s costs. 
• Accounting - A look into your potential customer’s accounting data allows for a short and long term view of how the business is doing. You can see at the transaction level where they are both taking in and spending money, and ascertain things like year-over-year growth and seasonality in the business. One drawback to this type of data is that it can be faked, though it is often possible to see timestamps and understand if large amounts of data were recently changed. 
• Banking - Viewing your customer’s banking data can give you real time insight into your potential customer’s income and outflow which is difficult to impossible to fake. The drawbacks are that you are only seeing the money go in and out with no broader context, and many bank APIs limit your lookback period to two years. 

Novel data sources allow financial institutions to see the same information you always needed–your potential customer’s financial health–but in new and robust ways.

The exciting thing about novel data sources is they allow you to take smarter and more calculated risks. Now, lenders can take on new customer segments and potentially even underwrite them on a daily basis.

"If you're in the business of saving time and money for other businesses, data has to be at the heart of everything you do." - Sri Srinivasan, Head of Risk at Ramp

But how do you make use of this firehose of data?

Designing Smart Onboarding Workflows with Novel Data Sources

The push today is for end-to-end automated underwriting that reduces the human-in-the-loop as much as possible. 

Srinivisan refers to these as the “three vectors of small business lending:

1. Minimal data - Only collect the data you need to make an underwriting decision. This will vary by customer and lending product type. 
2. Minimal friction - Strive to make onboarding simple for your customer. Marry the information you need for the lending product with the customer’s comfort level.
3. Minimal overhead - Streamline all data collected so you don’t overwhelm your underwriter.

Similarly, Prathik Naidu, Product Lead at Rutter, recommends thinking of your customer type and product type as a matrix. For example, if you are offering a line of credit and your customer is an e-commerce business, then using granular order level revenue and payout data can help you more thoroughly understand the financial health of their business.

Along with implementing novel data sources, a second tactic is to decision in stages to balance the risk vs experience trade-off. For example, you could go ahead and approve the aforementioned e-commerce customer based on their bureau data, but offer them better terms if they agree to share their financials through an integration. 

With this onboarding flow, lenders can get to an approval stage while still mitigating the risk associated with a new and unknown customer. 

According to Naidu, this layered onboarding flow has allowed lenders to update their service level agreements (SLAs) from a standard 2 days to, in some cases, instant. 

With this newfound speed-to-quote, modern lenders are finding it much easier to win.  

Strategically Deploy Friction, Gain Speed

Leading lenders are also strategically deploying friction to improve their speed to quote and balance the tradeoffs between customer experience and risk.

According to Abby Grills, Product Manager at Middesk, successful lenders are defining clear rules upfront and introducing risk-based workflows to automatically filter out the easy yes's and easy no's. These auto-approval and “knock-out” rules free up analyst time to focus on the applications that most need human attention. 

Of course, you need high quality data to confidently make these decisions. Poor data can lead you to reject good candidates or send bad ones on to an analyst. For example, a lender may automatically reject customers with liens. With legacy data sources like Dun & Bradstreet or LexisNexis, this can happen even if those liens have been terminated. But Middesk’s Liens search returns the most up-to-date lien data, ensuring your prospect isn’t filtered into an “automatic no” due to stale data. 

Another issue can arise because some data points, like industry, are self-attested. Middesk Industry Classification reveals a business's true industry to uncover if they are involved in an often-blacklisted industry like gambling, firearms or cannabis. 

Strategically deploying friction around automatic approvals and rejections streamlines your pipeline and frees up human analysts for more complex files.

Iterate, Iterate, Iterate

Sri Srinivasan, Head of Risk at Ramp, recommends first introducing a novel data source to your onboarding flow in a space where you are already rejecting many applicants. 

This tactic allows you to experiment with asking customers to share data in a way that isn’t disruptive to your current customer onboarding process. 

To do this, ask:

• What piece of data would help me innovate in this space?
• Is there a novel data source that covers that piece of data? 
• What does the new flow look like after introducing this request?
• Are customers comfortable giving me this data or do they drop off?
• Is it efficient?

Once you have taken some time to test introducing a novel data source here, you can then consider scaling it to other areas that may be more critical and where you are more hesitant to disrupt the flow of applicants. 

Ramp, for example, integrated 3rd party accounting data in an area where they were mainly rejecting applicants. So rather than disrupting applicant flow, accounting data allowed underwriters a second look at the potential customer and sometimes turned that “no” into a “yes.” 

How Lenders Can Adopt AI and Behavioral Data

The panel generally agreed that artificial intelligence will become a critical factor in improving lending speed in the next few months. 

Large language models are ideal for parsing and comparing disparate data sources, such as those collected from both traditional and novel data sources. “It would take a human three different screens and an Excel sheet,” said Naidu. 

Banks and other FI’s already use models to predict the future and inform sound decision making. The concept of “model governance” in banking requires that banks rigorously test that their models are used appropriately while also adhering to regulations around fairness in lending. Grills cautions that there could be a struggle to actually apply AI learnings and then justify why the model made the decision that it did. She raised that this could lead to a regulatory slippery slope if the decisioning behind the technology can’t be explained and justified.  

The panel was more circumspect when it comes to the use of behavioral science data. They agreed that behavior patterns, such as copy pasting large amounts of information or filling out an application in all capital letters, would be more useful indicators of fraud than actual signals when it comes to underwriting. 

Bets on the Future of SMB Lending

The use of novel data sources like accounting, banking and commerce payments data is about to revolutionize the way financial institutions underwrite. So what does that mean for the future? 

Jay Patel, Product Partnerships Lead at Middesk, bets that – with underwriting becoming quicker and more cost efficient – leading lenders will use the newfound capital to expand into new verticals. 

Srinivasan predicts that the future is embedded financial products, and the challenge is now to understand which novel data sources work for which customer at which point in their journey. 

Grills is sure that, now that small business lending looks like less of a risk, more non-traditional players will get into small business lending. This will allow for more opportunity and greater SMB growth. 

And Naidu predicted that we’ll soon see more than just the traditional small business financial products like lines of credit and term loans. Innovators using novel data sources will now be able to offer custom products and terms that dynamically and autonomously change over time to benefit both the lender and the business. 

To sum it all up: exciting times are ahead for SMB lending.

Click here to watch the webinar.

‍