Financial technology and lending decisions are built on data. Lenders can buy data in the form of reports, statistics, and bits of information about consumers and their buying habits from hundreds of companies and accumulate mountains of their own new data from every portfolio and process.
Employing certain practical ideas, lenders can get the most bang for their data buck and gain insight that can translate to real advantages.
Better Data from Exceptions
In almost any type of lending, better data starts within the lending business itself. Every lender has processes that create really valuable data. But for most lenders, valuable data leaks out of their processes throughout a loan's life. That's not hard to fix and it's not expensive -- it just requires a commitment to maintaining a rock-solid data foundation.
If you ask your rich uncle for a loan, maybe there aren't many rules. Even so, your uncle probably wants to know how you're going to repay the loan. Because your uncle likes you, he might give you the loan if your answer isn't great. Rich uncles can do that. Lenders should definitely not do that and they mostly believe that they don't. In reality, they do and they do it often. Manager overrides, manual exceptions, and virtually every other subjective decision when making a loan are almost always bad ideas and they muddy some of the most critical data a lender can acquire.
Override gains offset by defaults
Evaluation of numerous portfolios purchased and sold (by the analyst team at Newport Capital Fund, a hedge fund that specializes in acquiring and disposing portfolios of distressed loans) showed a consistent result - portfolios almost never get a true and sustainable lift in profitability from policy exceptions or manual overrides. In the short term, a statistical wash can happen, but that's about as good as it gets. A few extra good performers gleaned with an override are almost always offset by extra defaults. Lenders usually lose money on exceptions. Despite all that, there are indeed profitable ways to use exceptions.
Exceptions and overrides are fundamental components of the evolution of credit policies. They clearly indicate policy mutations. When lenders manage exceptions strategically, they quickly make those that cause losses extinct and just as quickly promote truly profitable policy exceptions to real business rules.
A decision to override a policy rule is always supported by a reason, usually based on experience. A typical scenario: "Teachers from Portland who buy Priuses always pay, so let's approve this one even though we require two years of residence history and this borrower only has 18 months." That choice to bypass a business rule is informed by experiential data from the manager's head. If it is the right call, it would be more useful as part of standing policy and implemented in the loan origination system.
The key is to do better than simply storing that knowledge as "Override - Other" in the loan origination system. Profitable lending policy evolution isn't a natural process - it requires some basic genetic engineering.
It starts with as robust a set of exception reasons as possible -- in an automated loan origination system or even in a spreadsheet if that's where the underwriting happens. That list is never too long. Not everything can be anticipated, so it's okay to use "Other" with one critical caveat: "Other" must have immediate review and the conditions that made it "Other" need to become lending policy going forward.
To achieve maximum data effectiveness, lenders must always be paring down their exception reason list. In a perfect data world, there would be none. All exception and override logic simply must be included in standard policy rules. True statistical validity, and thus, optimal portfolio performance, cannot be achieved if the rules allow and the data contains poorly documented or subjective decisions, even if they comprise a small set. When every lending decision uses all the data and logic available, with ongoing monitoring and quick adaptation, optimal lending policy evolves and the potential for profitability grows exponentially.