Hi and welcome back,
In this article series, I’m covering the key questions most challenging banks address on their way to profitable lending. Here are they:
Check out the first part of the article. There I provided some ideas on the first two aspects. In this text, I have a lot to say about points 3 and 4, so let’s discuss how to understand your market and your buyer persona and how to sketch a set of profitable lending terms and build a working loan management system.
In the upcoming third part of the article, I will provide some ideas on launching loan products as well as underwriting models and how to embrace the changes and finally start lending (profitably).
Probably, the easiest thing in theory and the hardest in practice: do you know exactly who forms your key customer base? How to take this data and calculate lending terms on its base? Why all these individuals (or businesses) should even become your borrowers?
You can find the answer in data harvesting and analytics. To ensure that the additional capital is used for good and to improve the customer’s life, a financial institution should determine who the beneficiaries are and which terms are beneficial, affordable, and feasible for the customer to repay.
The decision of whom to lend to might also depend on the banking strategy, including your risk appetite and long-term strategy. For instance, if a bank targets SMEs, it might offer niche products for Amazon sellers, which would make the lending proposition more consistent with other product suites.
Moreover, the question of whom to lend to also depends on the level of risk that the bank is willing to tolerate. Typically, banks that start lending may have a moderate risk tolerance, but it’s clear that not all portfolios will default.
If a bank lacks sufficient actionable data or expertise, we recommend starting with the lending marketplace business model. In this case, the bank originates loans on behalf of other parties, which is safer.
By operating as a lending marketplace, banks may gather essential data about their customers and analyze them from a risk perspective. The best thing about this business model is that banks can build more robust lending programs by leveraging the data and expertise of their lending partners. With the growing amount of collected data and expertise, you may start direct loan origination.
The lending marketplace model is a great starter and a nice augmentation of existing services.
In this case, you may want to start with testing business models with basic data providers and lending small amounts in short terms.
Sure, it is essential to keep an eye on your competitive edge from the very beginning. Adding more data providers, no mater alternative or classical will help you collect more user data to build lending patterns later.
While you may need your own advanced data collection, harvesting, and storage tools in the future, the main idea here is to start by collecting additional data through many small checks. Once you collected a sufficient amount of data for AI/ML, you can experiment with it to build smart financial risk prediction models that rely on these extra data points to enhance prediction, offerings, and financial outcomes.
In order to do that I’d suggest using the combination of IRR / WACC as a benchmark. Yes, it takes additional adjustments regarding risk, loan product types, lending market segments, or predicted operating expenses, along with customer segment affordability and market offerings.
To draw product terms, update the following data:
The customer segments affordability concept definitely requires attention, so I suggest that you made your own research on that (or simply check out this article).
Loan origination is maybe 20% of the entire lending cycle. Well, not if you’re the originator, but still there is one important thing to remember: having access to loan management history and borrower data, including information about default loans, is a must-have.
It is a must-have if you want to build a data-rich system for rapid and informative decision-making and improving repayment rates.
Due to PSD2 in Europe, you can access a whole lot of lending data. Also, you may ask the liquidity providers to include that in the agreement or set up your CRM to enable real-time insights or regular data extractions.
If you keep loans on your balance sheet, please, make sure loan management is set up in a very efficient way for timely debt collection and proactive work with delinquency.
Merely having a sufficient amount of data enables your ML team to identify effective notification patterns and proactive strategies. By collecting and analyzing data, you can understand which notification methods and actions work best with each customer segment and which loan products can generate the most revenue over the long term. You can also attempt to identify patterns in customer behavior to create efficient notification campaigns that reduce delinquencies and defaults.
Another crucial point is implementing proactive strategies. By utilizing data collected from previous customers and analyzing the behavior of your existing customer base, banks can anticipate customer behavior and take proactive measures to mitigate potential risks. These measures may include cutting credit limits, adjusting APRs, offering refinancing or payment postponement services, or negotiating with clients before their financial situation deteriorates and affects the bank’s portfolio negatively. The same mechanism can also be used to identify positive anticipated behavior, such as a second-in-a-year salary increase, which can be used to offer new loan products or adjust existing lines.
It is crucial to ensure that the collection options for end customers are as seamless as possible. Investing time and effort into integrating with modern payment processors and enabling features such as auto-payment or auto-invoicing for SMEs and 24-hour payment capability can enhance the customer experience and streamline the payment process.
As you may have noticed, data makes the difference. Obviously, the role of data and its implementation methods is something that we are going to focus on in the third part of this article.