By Max Penel, Investment Director & Co-Head of Global FinTech
FINTECH FUNDAMENTALS, PART 1 OF 2
The platforms that endure are defined not by how quickly they grow at launch, but by how deliberately they incorporate learnings and adapt. After funding and diligencing fintech and embedded-lending platforms across 15+ countries and 12+ years, that pattern recurs with striking consistency. Those that scale sustainably execute the basics exceptionally well, and what separates great from merely good comes down to two things: a distribution edge and a data edge. Either they can originate at a lower cost and better economics, price risk more accurately than the competition, or both. Platforms with both advantages tend to compound quickly.
Data Is the Foundation, Not an Afterthought
Without clean and reliable data, fintech platforms cannot learn, iterate, or improve. And without that learning, underwriting, pricing, and collections never materially improve. Lacking a strong data foundation, platforms cannot accurately assess or price risk, attract capital from sophisticated lenders who require transparency, or build a durable edge in underwriting or distribution.
The subtler danger is false confidence. Poor data doesn’t announce itself; decisions look rational based on incomplete signals, then fail at scale when the gaps become visible. By the time the problem surfaces, it has usually compounded.
Internal discipline is the starting point. A fit-for-purpose loan management system, proprietary or third-party, it largely doesn’t matter, must capture data accurately, consistently, and at sufficient granularity. Not monthly summaries, but repayment behavior at the transaction level, timestamps, and cohort-level performance that can be sliced and interrogated quickly.
That internal foundation is necessary but not sufficient. Access to third-party data, bureau feeds, telco signals, and embedded-platform behavioral data, further sharpens underwriting and widens the gap between great and good. Platforms that combine clean internal data with high-quality external signals tend to price risk more accurately and improve faster than those that rely on either alone.
Focus Only on Where You Truly Add Value
The most effective fintech platforms are ruthless about focus. That starts with an honest assessment of where the company genuinely creates value, and equally, where it doesn’t.
The strongest platforms concentrate capital and leadership attention on the capabilities that define their competitive edge, whether that’s underwriting speed, distribution reach, or customer experience. Everything else is a candidate for outsourcing, de-prioritization, or cutting entirely. Building technology that doesn’t directly enhance the core value proposition is a distraction, and at any stage, a costly one.
The temptation to broaden is persistent and usually dressed up as optionality. Resisting it is one of the harder disciplines of early-stage platform building, and one of the most consequential.
The platforms that endure are defined not by how quickly they grow at launch, but by how deliberately they build understanding.
Early Losses Are More Harmful Than Most Founders Appreciate
Losses incurred early in a platform’s life can be particularly damaging, but their nuances and details behind those losses matter more than their presence. For concentrated or larger-ticket portfolios, even modest early losses can distort underwriting signals and erode lender confidence. For smaller-ticket, high-velocity books, some losses are expected and manageable, provided they are consistent and understood. What distinguishes problematic early losses is their unpredictability: lumpy, inconsistent, or uncontrolled losses signal weak foundations rather than normal credit absorption.
The most consistent failure point we see is when companies have not solidified product-market fit but still expand into larger balances or longer tenors to chase volume which ultimately compounds their exposure. Without a sufficient equity cushion, debt funders will pull back or restrict capital as loss patterns deteriorate – and recovery from that position is rarely straightforward.
The platforms that survive run controlled experiments. They A/B test on small cohorts, back-test changes on broader datasets, and accept slower growth initially in exchange for systematic implementation of their learnings. That discipline pays forward, cleaner data attracts better capital, on better terms, and earlier in the platform’s lifecycle.
Product Design Determines Unit Economics
For most fintech lenders, the largest cost lines are customer acquisition, cost of funds, and credit losses, with collections costs often close behind. The strongest platforms design products that structurally reduce all of them.
Buy-now-pay-later used merchants as distribution channels, dramatically lowering customer acquisition costs. Embedded-finance models go further: cash dominion reduces both loss severity and collections friction simultaneously, while distribution is embedded in the platform itself. When product design addresses unit economics, margin expansion is structural and scale becomes a function of execution.
New Platforms Will Be Targeted
Fraud is not an edge case for early-stage fintechs. It is an absolute certainty. New platforms are attractive targets precisely because they lack seasoned transaction history, rules engines aren’t calibrated, anomaly detection has no baseline, and organized fraud groups know exactly when and how to exploit that window. Account takeovers, synthetic identity clusters, and coordinated application attacks often appear within months of launch.
The platforms that survive think adversarially from day one. This means someone in the business, not a compliance function, but a risk or product owner, is actively modeling how the system can be gamed: velocity patterns that signal synthetic identities, device fingerprinting gaps. The question is not just how the product should work under normal conditions, but how it breaks under deliberate pressure.
Short Feedback Loops Build Better Credit Businesses
The faster customers repay, the faster platforms learn. Shorter tenors and higher-frequency repayments accelerate feedback on underwriting quality, collections performance, and customer behavior, compressing the iteration cycle that would otherwise take years into months.
Shorter feedback loops also allow platforms to demonstrate collectability to funding partners far earlier in their lifecycle, directly improving the terms and size of capital they can access. Longer tenors delay all of this. By the time underwriting problems surface in a long-tenor book, they have usually compounded, and the cost of fixing them, in lost capital access as much as credit losses, is rarely small.
Fewer Inputs, Stronger Signals
Early underwriting models tend to over-rely on inputs with weak predictive power. This is usually a volume problem: without sufficient transaction history, platforms compensate by collecting everything in the hope that something correlates. The result is models that increase application friction without meaningfully improving credit outcomes.
Improving signal quality is a function of time and seasoned data, or access to high-quality external sources for platforms that lack portfolio history. But the more counterintuitive discipline is removal: the platforms that improve acceptance rates and reduce losses are often those that aggressively cut low-signal variables and concentrate on the handful that actually correlate with repayment behavior.
Fewer inputs, stronger signals. The simplest models, built on the right variables, tend to outperform complex ones built on many weak ones, and they’re easier to monitor, explain, and improve over time.
This is Part 1 of a two-part series. Part 2 covers understanding risk toggles, managing product expansion, and building a capital runway to sustain growth.
—
The views expressed are my own and do not necessarily reflect those of my employer.
This content is for informational purposes only and does not constitute an offer to sell or a solicitation of an offer to buy any securities. Any such offer will be made only to qualified investors through confidential offering documents. All investments involve risk, including the possible loss of principal. Past performance is not indicative of future results.
