Cathy is the owner of a small organic ice cream parlor in Munich. She is 25, and the store, her first business venture, has been in business for just over a year. She has relatively few substantial assets: the display chillers where the ice cream is kept, the ice cream makers, a good espresso machine, and all the associated bits and pieces necessary to make things work. She enjoys a steady flow of customers, but revenues are naturally somewhat seasonal: sales drop in the winter and spike in the summer. Aside from a few final payments on the equipment, she’s almost at break-even.
Optimistic about potential, she’d now like to open a second location and needs funds for outfitting the store, marketing it, and purchasing initial inventory.
At first blush, Cathy’s business looks like a risky, challenging case from a lending perspective; any risk officer would say as much. It’s only been open a year, sells a niche product, has seasonal dips, and offers few collateral assets. The problem is, of course, that Cathy needs a loan.
It is the key element of every business loan process: the lender must assess a potential SME borrower's business and assign a risk level to determine first the binary decision of whether or not to lend, and then amount, price, and repayment terms of lending.
Even before deep-diving into the company’s credit rating, current financials, and other indicators, there’s one initial factor that can hold significant weight in the scoring process: bank customer segmentation.
A risk segmentation approach essentially makes broad assumptions, identifying applicants in categories that historically and intuitively indicate higher or lower risk. While every lender has its own criteria, most lenders are wary of businesses that:
● Have no track record, even if lead by a proven and experienced team;
● Can be considered startups, headed by young or inexperienced leaders;
● Are able to offer no physical assets as collateral or have insufficient collateral;
● Come from industries like gaming, extraction, crypto, or non-standard pharmaceuticals that are highly regulated (regulations change, can hamper growth, and even unintentional violations can ruin a business);
● Traditionally represent businesses with high levels of risk;
● Rely on manufacturing non-essential products, without an existing client base, as these involve considerable overhead;
● Struggling with cash flow (ironic, yet understandable from the bank’s perspective) — regardless of the reason why;
● Carry high debt-to-income ratios.
For many lenders, any one of these may be enough to turn down an SME loan application or at least to relegate it to the bottom of the pile. For most, a combination of these will certainly do so.
Also, bear one more reality in mind: this approach to customer risk segmentation applies when business conditions are benign. For example, new factors and risks came into play during the COVID pandemic; the entire hospitality, travel, and entertainment industries temporarily (and logically) joined this list of high-risk borrowers.
Why Is Risk Segmentation a Problem?
No financial expert will ever argue against using all available tools for credit risk management in respect of a lending business.
A lender's role, for the sake of profitability and compliance with regulator policy, is to determine, as best as can be, which industries or business profiles represent a higher likelihood of defaulting on a loan.
As such, using an initial, high-level customer segmentation approach to eliminate general types of potential clients immediately reduces the overhead required to do manual assessments of financials and other factors that might lead to rejection.
Done correctly, customer segmentation strategies also allow a lender to proactively market to specific segments that it deems most creditworthy. It can do so based on data of prior performance of financing this segment, with messages that resonate with those borrowers: “Because of who you are, your application has a higher-than-average chance of being approved.”
However, it’s not all positive. Along with these clear benefits comes the well-known drawback of any type of high-level profiling, or use of unquestioned “traditional” criteria: It neglects the Long Tail effect — the very long list of SME borrowers who, although falling into a broad but classic higher-risk category, may be nonetheless fairly safe for other reasons. In credit circles, this is referred to as reject inference.
Segmentation Repurposed as a Drill-down Tool
As mentioned, segmentation in banking is a handy mechanism because it removes a lot of unnecessary manual work later on. However, its one-size-fits-all approach makes wide-ranging assumptions that eliminate potentially profitable business. What if we were to radically change that next phase by eliminating the expensive, time-consuming manual work in those later steps, thereby reducing the need to segment early on and exclude large numbers of prospective borrowers?
This is precisely what the newest embedded banking platforms can offer.
In the old banking paradigm, an applicant would have to submit substantial amounts of paperwork (assuming it exists) in order to validate (and increase) eligibility for a business loan. Newer technologies allow banks to automatically and near instantaneously collect this data via APIs and other data-sharing technologies, hundreds of variables at a time. This approach allows the lender to dig deep into the company’s background without needing that first generalized opinion - often developed in the absence of sufficient evidence - on a segment or category.
With this approach, the lender may discover new patterns and realities that don’t align clearly with traditional common wisdom or older business realities. Recalling the segments above that had the potential to immediately eliminate an SME’s application, a bank could potentially learn that:
● Age may have little or nothing to do with success in specific industries; the mindset of veteran experienced leaders may be balanced out by the passion and energy of younger founders;
● A service industry without collateral may have its unsecured status balanced out by the fact that it also requires less overhead and can secure long-term contracts ensuring that cash flows can be assured for a period of months or years into the future;
● Specific industries in specific geographical locations may have advantages due to a particularly loyal/relevant customer base;
● Second-level categorization, like specific types of food service businesses or specific localities for commercial real estate improve risk assessment instead of being swallowed up by the top-tier segmentation category.
These are all hypothetical learnings, but as you can imagine, the list is near-infinite once the lender begins to leverage granular data that can create clear patterns (both positive and negative) with much, much deeper levels of risk segmentation. Not only can this strategy open the door to previously rejected applications – it may quickly identify micro-segments that are specifically safer, and therefore worth seeking out. In short, it drives new business.
In Cathy’s case, for example, the drawbacks – including her age and short business track record – could be more than offset by other considerations. For example: the fact that organic ice cream is only growing in popularity in her community, and she is the exclusive seller; her profit margins are high; as a single woman with few personal expenses she pays herself a modest salary and works late hours; she is a social media guru and drives business through using it as a channel. Few of these considerations are measurable or would have been included in a traditional assessment. However, her product category, location and age were clues that could indicate a trend for the bank to look for in other applicants.
Flexibility is the Secret Sauce
The ability of embedded banking solutions to quickly collect data and make a more specific and situational assessment is only the first step. The real secret to success in this strategy is the agility and rapidity of testing and experimentation.
How would that work? Let’s say that one of the learnings above appears to support a change in a bank's credit policy towards a particular micro-segment. The next step is to test the theory in three basic stages:
1. Create a rule that takes into account mitigating factors and allows specific business types that would have been rejected in the past to be now accepted, perhaps with specific terms.
2. Create highly targeted marketing campaigns to draw these potential applicants to supplement the natural flow already approaching the bank.
3. A/B test the concept by setting up a control group, and monitor both sets of borrowers.
Naturally, there are two problems with this process, each of which is solved by the flexibility of quickly changing rules: First is an increase of credit risk through this experimentation; second is the time it takes to run these tests, waiting to track long-term default statistics. The more slicing and dicing, the more work it is to configure and track.
Let’s first address the fear of introducing new credit rules that ease lending without increasing risk. Today’s fintech solutions solve this by limiting any potential downside: allowing only a limited number of carefully defined applicants to receive loans based on specific criteria. For instance, a bank may want to test the theory that young, aggressive business owners in service industries with offices in the suburbs are low risk because of their lack of overhead like expensive equipment and real estate. In addition, this demographic’s use of the newest technologies may give them the edge against traditional competitors. A campaign targeting only these business owners can send them to a specific rule flow in which their loan application must include these traits. Only 50 are accepted and closely tracked. In parallel, a similar group of new service-oriented businesses with few expensive assets – but without the age or location factors – can be accepted as well. Comparing the payment/default patterns of both groups can indeed indicate whether youth and location increases or decreases risk – or makes no difference.
Credit risk is controlled, while rapid tracking (especially for short-term loans) saves time.
In short: Without making this a comprehensive, global change, the ability to quickly change, add or remove rules, essentially reduces the risk of this valuable experiment. Add to this the option to create any type of granular test in just minutes, and the scope — and, thus, financial upside — of this optimization is massive.
ezbob’s Flow Manager is designed to streamline these processes. Where in the past a bank might take days or weeks to adjust a risk policy, ezbob clients do it in minutes. The ability to create as many micro-categories as desired — all in a low-code, quick-click manner — changes the entire strategy of risk assessment optimization. What makes this all even more exciting to our clients is that after a deployment that, from the start, requires few changes by the bank’s IT team, all these configurations and rule adjustments happen on the ezbob platform, without changes to the lender’s physical or data logic infrastructure.
To summarize: Customer segmentation in itself is a critical and valuable tool for SME risk assessment, but as we’ve seen, it’s also a way to reduce the number of valid business loans when the filtering process is too broad. Fintech providers such as ezbob offering embedded lending platforms provide solutions with granular analysis that “understand” an applicant’s business more thoroughly, opening the door to both more effective, targeted marketing and more completed transactions.
Want to learn more on our embedded lending solution, head to our platform page.