Latest / Fintech Conversations with Fexingo: Banking, Payments, and Financial Software Companies / How Biometric Lending Redefines Credit Access
Transcript
- Lucas: The credit score has been the bedrock of American finance for decades, but as we head into the final quarter of twenty twenty six, it feels increasingly like a rearview mirror. I have been digging into how fintechs are now moving beyond static FICO numbers to use what they call dynamic behavioral signals. The idea is simple on paper: instead of looking at what you did three years ago, look at what you are doing right now. Luna: Right, because a FICO score tells you if you were reliable in the past, but it does not necessarily tell you if you are going to pay back a loan next Tuesday when your inventory turns over. So what specific data points are replacing that old metric? Lucas: It is less about one magic number and more about a constellation of signals. We are talking about transaction velocity, which measures how quickly money moves in and out of an account, alongside device stability and even typing cadence during application processes. For example, a merchant who consistently processes payments before noon every day shows a level of operational discipline that a generic income statement might miss entirely. Luna: That sounds incredibly granular. Does that mean two people with identical incomes could get wildly different loan terms based on how they type or when they open their banking app? Lucas: Exactly. And that is where the ethical tightrope gets very narrow. One fintech I spoke with last week noted that their approval rates for gig economy workers jumped by thirty percent when they switched to this model. But they also saw a slight uptick in fraud attempts because bad actors started trying to mimic those specific behavioral patterns. Luna: So the system is learning, but so is the fraud. How do they stop someone from just acting 'responsible' for ten minutes to get approved? Lucas: They don't rely on a single session. They build a longitudinal profile. If your typing speed is normal but your mouse movements are jittery, or if you apply for a loan at three in the morning after being inactive for weeks, the algorithm flags it. It is essentially building a digital fingerprint of your financial personality rather than just your bank balance. Luna: That raises a huge question about accessibility though. Not everyone has a smartphone or consistent internet access, which means this model might inadvertently exclude the very people it claims to help. Lucas: That is a fair critique. However, in markets like Southeast Asia and parts of Africa, mobile penetration is high enough that this data actually exists for people who have no formal credit history. In the United States, we are seeing a slower adoption because consumers are more skeptical about sharing this level of behavioral data. Luna: Skepticism is healthy. When you give up your typing patterns or location data, you lose a layer of anonymity that used to be guaranteed by just having a checking account. Lucas: Precisely. The trade-off is visibility for access. You become visible to the lender, but you also gain access to capital that was previously locked behind a wall of traditional documentation. It is a fundamental shift in the power dynamic of lending. Luna: I think many small business owners would take that trade-off if it meant getting funded in hours instead of weeks. Speed is everything in retail. Lucas: Absolutely. Let's talk about that speed. A traditional bank might take forty-five days to underwrite a small business line of credit. These new platforms can do it in under ninety seconds. That is not just a convenience; it is a survival mechanism for businesses operating on thin margins. Luna: Ninety seconds is impressive, but does that speed come with higher interest rates? Usually, faster decisions mean less due diligence, which implies more risk premium. Lucas: Not necessarily. In fact, some data suggests the opposite. Because the risk assessment is more accurate, lenders can offer lower rates to low-risk borrowers who were previously lumped into the average bucket. It is price discrimination based on precision rather than prejudice. Luna: Price discrimination based on precision is a much nicer way to put it than price discrimination based on zip code. So this is actually a step toward fairness? Lucas: Potentially, yes. If the algorithm can ignore demographic proxies and focus purely on repayment behavior, it removes human bias from the equation. Of course, you still have to worry about algorithmic bias built into the training data itself. Luna: Which brings us back to the data quality issue. If the historical data used to train these AI models is biased, the output will be biased too, regardless of how fast the decision is. Lucas: That is the industry's biggest headache right now. Regulators in the European Union and several US states are starting to demand explainability. They want to know why the bot said no. You cannot just say 'the computer said so.' You need a reason. Luna: Explainability is hard when you are dealing with millions of data points. How do you translate 'typing cadence' into a legal justification for denial? Lucas: You generally don't cite the typing cadence directly. You cite the resulting risk score. But that creates a black box problem. If the borrower doesn't understand why they were denied, they cannot fix it. They just move to the next lender, who probably uses the same model. Luna: So we end up in a loop of rejection without improvement. Is there any company currently solving this transparency issue? Lucas: A few startups are experimenting with 'feedback loops' where they tell you exactly which signal hurt your score. If it was a late payment, fine. If it was erratic spending, they might suggest a budgeting tool to stabilize it before you reapply. It shifts the relationship from adversarial to advisory. Luna: That is a fascinating pivot. Instead of just judging you, they try to coach you. It feels more like a partnership than a gatekeeping exercise. Lucas: It is. And that is likely where the market is heading. Banks are realizing that keeping customers for life is cheaper than acquiring them once. By helping you manage your risk, they increase your lifetime value. Luna: But do they really care about your lifetime value, or just your current default probability? Those aren't always aligned. Lucas: Good point. There is a tension there. But for now, the efficiency gains are driving adoption. I am seeing more traditional banks partnering with these fintechs to integrate these behavioral layers into their own legacy systems. Luna: So the big incumbents are buying the agility of the startups. It seems like the old guard is finally admitting that their own data is too slow. Lucas: They are. And it is forcing a rethink of what constitutes collateral. In the past, you needed a house or a car. Now, your data trail is becoming the asset. Luna: Your data trail is the collateral. That sounds dystopian and liberating at the same time. Which side wins out? Lucas: For now, liberation wins for those who have the data to prove their reliability. For others, it is just another barrier. But the trend is clear: static scores are dying. Luna: And if you're not part of the digital conversation, you might not get a seat at the table. It is a brave new world of finance. Lucas: Indeed. As we close out this quarter, watch for regulations that force these algorithms to be more transparent. The era of the invisible score is ending.