The Risks to Financial Inclusion

James Yagley
Founder and CEO, Fafnir Lab
February 16, 2022

I’d like to begin by thanking Garrett Menning for the invitation to contribute this guest blog on Black Pearl’s site. Garrett asked me to discuss my company, Fafnir Lab, a fintech company based in Detroit, and share our perspectives on financial inclusion. Financial inclusion is too broad a topic for one post, so permit me to focus on the specific challenge of access to credit for small businesses.

Fafnir Lab’s approach is to combine rigorous data science with a sensitivity to how people think and act in the real world. We call this latter dimension “cultural practice.” In the field of data science, there is an explicit recognition of the need for what they call “domain knowledge” to ensure that mathematical models will work in production (i.e., in the real world). In the language of the business world, we sometimes call this “customer empathy” or “meeting the customers where they’re at.”

These are different expressions of common sense, the kind of sentiments that most everyone can agree with. Unfortunately, over my career, customer empathy has been the exception, not the rule.

Again, our work involves enhancing small business/SME access to credit. Our focus is on lending to meet operational expenses or expansion. Credit involves risk. This is a truism for enterprises of all sizes, in all times and places. In conducting the work of expanding access to credit in underserved markets, I’ve often heard the “truism” that “lending to underserved populations is the riskiest kind of lending.”

Is this the reality? Or a perception? It’s not enough for us as practitioners to embrace an alternative point of view. We need to develop credit models that can work in production – again, in the real world. We start this process by going back to the fundamentals.

One of them is the distinction between risk and uncertainty.

Risk versus Uncertainty

In everyday language, we often use these words interchangeably. But they have two different – diametrically opposed – meanings in finance. Both mean situations where something goes wrong. A borrower defaults on a loan. An investor loses his or her capital.

Such an adverse event is uncertain if the odds of it happening can’t be measured. With uncertainty, you are completely in the dark.

Risk, on the other hand, means we can somehow measure a negative event. In lending or investing, we often measure risk in terms of probability, or odds.

Risk is still something we want to avoid. But it is immeasurably better to be dealing with risk rather than uncertainty. Whether it’s high or low, risk is something you can plan for. Prepare for. You can put structures in place around risk to defend against it.

This is where data can make a difference, of course. Having good quality data (or knowing how to look at data) can turn a situation from uncertainty to risk. It can make the measurement of risk more precise. Potentially, it can even show there is far less risk than we first assumed.

Risk is good—or rather, moving from uncertainty to risk is good. This is what makes economic exchange possible.

Financial inclusion is not a challenge only in the developing world. In the US, we’ve seen increased recognition of large segments of the population who are “credit invisibles.” These are populations who, for one reason or another, are unable to be scored by the traditional credit scoring agencies. Without a (risk) score, the lending outcome is uncertain. As a result, the “credit invisibles” are unable to access home loans, auto loans, and small business loans.

The Consumer Financial Protection Bureau finds a staggering 1 in 10 Americans are among the credit invisibles. That’s 26 million consumers. Specific sub-populations include young adults without a credit history, immigrants, and minority groups who have traditionally been excluded from full participation in the financial mainstream.

With today’s data revolution, there’s hope to change this situation. Mainstream financial institutions (banks) and nontraditional lenders all are developing solutions to evaluate the underlying financial condition of a credit invisible applicant, to determine ability to repay, probability for loss, and other factors, using a mix of different data sources and different machine learning models.

The end goal: a systemic approach to transforming uncertainty to risk.

The Limits of Data and the Importance of Customer Engagement

Fafnir Lab is a data science-centric organization, so it’s no surprise we see data playing the lead role in transforming uncertainty into risk. What might surprise you is how keenly aware we are of the limitations of data. Especially raw data or data that is one-sided in its interpretation.

We’ve come to recognize that high-quality data results from the act of communication. It flows out of a relationship. It strengthens connections between two parties so that their exchange brings greater value to both sides.

It seems like we are swimming in data, inundated in all directions from multiple sources. Companies market data to us as a raw commodity. You can purchase it in bulk, these marketers say, and watch your profits grow.

Isn’t this what we want? Isn’t increased data solving the credit invisibles problem? Classic communication theory reminds us that the task is to distinguish information from knowledge. I would say with all these new streams of data, we tend to forget that information, by itself, does not equate with knowledge.

Let me use an example from Fafnir’s current work: In the small business lending space, companies offer solutions that center on data integrations through application programming interfaces (APIs). At its heart, this is what Open Banking is about, using APIs to provide access to bank account data. There is a larger world beyond APIs, and we can access data from eBay or Etsy sellers or dozens of other online marketplaces to analyze a small business’ financial condition.

Often, when people talk about a data revolution, they are speaking about API pulls and data aggregation, which in turn feed into data science.

The temptation with APIs is to simply flow the data right into a machine learning model, applying the hottest software package without regard for domain expertise or understanding the subject’s challenges and circumstances.

One use case for API pulls is to analyze a small business’ cash flows. We agree this is an exciting potential application for this type of data. It’s one we’re pursuing as well. Our efforts reinforce the need to ask: How does the data sourced from APIs align with the business’ current cash flow dynamics? Their liquidity needs? Is data alone telling us the whole story?

Let’s say the data shows a healthy pattern of recurring monthly revenue, maybe even a recent uptick in sales. Is the incoming revenue needed to pay back trade credit for inventory purchases (in other words, to cancel out a pre-existing liability)? Are the funds applied to an outstanding receivable (transforming accrual income into actual cash)?

Perhaps it’s the opposite: the business operates a successful pre-paid model. In this case, the incoming funds will trigger new expenses.

Without an understanding of the business’ circumstances, a simple model may produce a very different picture of cash available to pay debt service, with significant consequences for credit seekers.

On this same example, one fallacy I’ve seen repeatedly these past two years is the equation of reduced expenses with business survival. Yes, in the COVID economy, business owners did need to reduce expenses to balance out their lost sales.

But for business owners, expenses mean productive activity. Wages pay for the people who build your product or sell your service. Increased wage expense is a response to increased demand. So is an increase in inventory expense or receivables.

Accounting for each of these nuances in the customer’s experience is an impossible task for a data scientist working in isolation. The necessary understanding of the customer comes as the product of a relationship. That’s why Fafnir engages with small business lenders and entrepreneurial support organizations. We want them as partners in our efforts. Their relationships with their clients, their customers, helps us complete the picture.

With this approach, small businesses gain value from the data relationship as well. The outcomes are improved business performance, expanded access to credit.

This probably sounds too easy … and it is. Establishing and growing a relationship is the hardest part for a company like ours. And when we reach that goal, the data journey is just beginning.

Who Controls the Data?

What is the alternative? Underperforming models that have no hope of achieving breakthrough?

A worse outcome than even this is one-sided control of data: privileged access to insights and actionable guidance from data.

This is the dark side to a system that uses data and algorithms to allocate credit. One-sided data means one-sided growth in value. Unscrupulous lenders charging exorbitant interest rates to a population that lacks choices. And doing it in the name of “serving a high-risk market.”

Risk becomes a dirty word again.

The Promise of Data Science for Promoting Financial Inclusion

Let’s end this on a positive note. There is genuine momentum behind expanding financial inclusion, and broad recognition of the transformative power that data and data science can play. The data revolution is just beginning. To transform this information into true knowledge, we need to improve our ability to develop relationships with SMEs – the people we’re trying to serve – to better learn from them, understand their contexts, and meet them where they’re at.

I think of this as relationship-driven data. It’s the human element that makes the data science work in the real world.

James Yagley is Founder and CEO of Fafnir Lab, a fintech based in Detroit that uses data science to meet the needs of underserved small businesses and consumers. He is also an Advisor at Black Pearl Consulting and Research.