There is a version of the fintech story that is told at conferences, in funding announcements, and in the opening slides of every neobank pitch: the incumbent banks were slow, expensive, and indifferent to ordinary customers. Technology disrupted them. Consumers won.
This version is not false. It is incomplete in ways that matter enormously — particularly for anyone who uses a fintech product and believes they are the customer rather than the product.
The Zero-Fee Illusion — What Free Actually Costs
When Robinhood launched commission-free trading in the United States, it was celebrated as a revolution. Within three years, it had introduced tens of millions of retail traders to equity and options markets. The narrative was democratisation.
The business model was payment for order flow — PFOF. Robinhood routed customer orders not to the exchange that would give customers the best execution price, but to market makers who paid Robinhood for the privilege of being on the other side of those trades. The market makers profited from the spread between the price customers received and the price available in the open market. Robinhood received a payment per order. Customers received a commission-free trade at a slightly worse price than they could have obtained elsewhere.
The fee did not disappear. It was restructured so that it was invisible, borne in execution quality rather than charged explicitly, and impossible for the average retail user to calculate or compare.
This model — eliminate the visible fee, monetise the invisible behaviour — is not unique to Robinhood. It is the architectural template of modern fintech.
Buy-now-pay-later platforms charge merchants a fee of 2-8% per transaction — significantly higher than credit card interchange — while advertising to consumers that the service is free. The merchant fee is embedded in product pricing, meaning consumers who do not use BNPL subsidise those who do. The BNPL user is not getting a free service. They are getting a deferred fee hidden inside the retail price of whatever they purchased.
Neobanks offer free current accounts and earn revenue through interchange fees on debit card transactions, premium tier upsells, and — increasingly — selling anonymised transaction data to financial institutions, advertisers, and credit underwriters. The account is free. The behavioural and financial data generated by using the account is not.
The Controversial Reality: Fintech Lends Most Aggressively to Those Who Can Afford It Least
The most persistent claim of fintech lending — from digital NBFC platforms in India to challenger banks in Europe — is that algorithmic underwriting democratises credit. Machine learning models can assess creditworthiness beyond the narrow lens of credit bureau scores, reaching the underserved and the underbanked who were previously excluded from formal credit.
This is technically accurate. It is also, in practice, a mechanism for extending high-cost credit to people whose exclusion from the formal banking system was, in many cases, a form of protection.
A person excluded from a bank loan at 11% interest who can now access a digital lending platform at 28-36% annual interest has gained access to credit. Whether they have gained access to a financial service that improves their position is a different question — one that fintech marketing materials do not address and regulatory frameworks have only recently begun to examine.
The algorithmic model optimises for default prediction, not for borrower welfare. A loan that is statistically likely to be repaid — even at high cost, even by a borrower who is structurally stretched — is a good loan by the model's definition. The downstream effects on that borrower's financial health are not a variable in the objective function.
The incumbent banks were exclusionary and often explicitly discriminatory in their lending practices. Fintech platforms are inclusive in access and frequently predatory in pricing. Both outcomes serve institutional shareholders more than the customers the platforms claim to champion.
This is not an argument against fintech lending. It is an argument for reading the APR before celebrating the approval.
What the Next Wave Is Actually Building
The first wave of fintech was distribution — building better interfaces on top of existing financial infrastructure. Mobile banking, digital wallets, commission-free brokerages. The underlying rails were unchanged. The experience was improved. The extraction model was refined.
The second wave, currently underway, is infrastructure — and it is significantly more consequential.
Embedded finance is the integration of financial services directly into non-financial platforms. A logistics company offering working capital loans to its driver network. An e-commerce platform underwriting inventory financing for its merchants. A payroll software provider offering earned wage access. In each case, the financial product is delivered by a company whose primary business is not finance, using banking infrastructure accessed via API.
The data advantage in embedded finance is unprecedented. A logistics platform knows exactly how many trips a driver completed, their income consistency, their seasonal patterns, and their equipment maintenance costs before it issues a single rupee of credit. No credit bureau has this data. No traditional underwriter can access it. The embedded lender's model is more accurate, lower cost, and — for the borrower — more accessible.
The counterargument, which is rarely made loudly in fintech circles, is that the same data advantage that improves credit access also enables a level of behavioural and financial surveillance that has no historical precedent in consumer finance. The embedded lender knows not just whether a borrower can repay — it knows how to structure repayment to maximise extraction while minimising formal default. These are not the same thing.
Central bank digital currencies — CBDCs — represent the endpoint of this trajectory. A programmable currency issued directly by a central bank, capable of being conditionally spent, automatically taxed, time-limited, or restricted by category. The financial inclusion case is real. The surveillance infrastructure implications are equally real, and are being discussed primarily in technical working groups rather than in public discourse.
The India Angle — UPI and the Infrastructure Paradox
India built one of the most impressive financial infrastructure achievements of the last decade in UPI — a real-time, interoperable payments rail that processed over 13 billion transactions in a single month in 2024. It is genuinely world-class public infrastructure, built by NPCI, that dramatically expanded financial access for hundreds of millions of people.
The paradox is that UPI's success has created a winner-take-most dynamic that concentrates transaction volume in two or three private platforms — PhonePe and Google Pay together process roughly 85% of UPI volume — while the economics of the rails themselves are structurally unviable for most participants.
Zero MDR — the regulatory decision to eliminate merchant discount rates on UPI transactions — means that the payments infrastructure operates at a loss for the companies processing those transactions. The cost is subsidised by venture capital, by cross-selling of adjacent financial products, and increasingly by the hope that the transaction data generated by UPI volume will become the underwriting foundation for lending, insurance, and wealth products.
The public infrastructure is being used to build private data moats. The beneficiaries of those moats are not the 500 million UPI users. They are the platforms that sit between those users and the rails — and the financial institutions that will eventually purchase access to the behavioural profiles those platforms have assembled.
This is not a conspiracy. It is a business model. It is worth understanding clearly.
SB Research Findings
Two structural observations from tracking fintech product economics and Indian digital lending data.
Finding 1: The Effective APR on BNPL Products Exceeds 40% Annually When Penalty Fees Are Included
The headline rate on most buy-now-pay-later products in India is zero — for the interest-free tenure of typically 15 to 45 days. The effective annual percentage rate, calculated by including late payment fees, processing charges, and the cost of mandatory insurance products bundled at checkout, consistently exceeds 40% annually when a borrower misses a single payment cycle.
The zero-interest framing is technically accurate for borrowers who repay within the tenure. For the segment of borrowers who do not — which, across BNPL platforms, ranges from 8-15% depending on the credit tier — the product is among the most expensive forms of consumer credit available in the formal market.
The disclosure of this effective rate is not required in the same format as traditional loan APR disclosures. It is fragmented across fee schedules, terms and conditions, and insurance product documentation. The average borrower at the point of checkout has no practical ability to calculate the cost of a missed payment before they commit to the transaction.
Finding 2: Fintech Lending Platforms Reject 60-70% of Applicants While Approving the Highest-Risk Segment at Premium Pricing
The democratisation narrative implies that fintech lending approves borrowers that banks would reject, bringing the underserved into formal credit. The data shows a more complex picture.
Digital lending platforms in India reject between 60 and 70 percent of applicants — a rejection rate comparable to or higher than traditional NBFCs for the same income segments. The approved segment is not the underserved middle — it is bifurcated. Prime borrowers with strong digital footprints receive competitive rates. The genuinely thin-file or new-to-credit segment, when approved, receives credit at rates between 28 and 42 percent annually — priced for high default risk.
The platforms are not wrong to price this way. The default risk is real. The question is whether the product being offered to the highest-risk approved segment — high-cost, short-tenure, auto-debit — is a financial inclusion tool or a mechanism for generating high-yield assets from borrowers with limited alternatives and limited financial literacy.
The answer, in most cases, is both simultaneously. That is precisely what makes it worth examining carefully.
The Question Nobody Asks Loudly
The standard measure of fintech success is scale — users, transaction volume, loan book size, GMV. These are the metrics in pitch decks, earnings calls, and press releases.
The metric that is almost never tracked publicly is net financial health of the user base. Are the people using these products — the BNPL borrowers, the high-APR digital loan recipients, the retail options traders on commission-free platforms — better off financially after two years of engagement than comparable people who did not use these products?
This data exists internally at every major fintech platform. It is not published.
The industry's argument is that access is inherently valuable — that having the option to borrow, invest, or transact is better than not having it, regardless of outcome. This is partially true and entirely convenient.
The next decade of fintech will be shaped less by product innovation and more by whether regulators, consumers, and the platforms themselves decide that access without outcome accountability is sufficient — or whether the standard shifts to whether the products actually improve the financial lives of the people they claim to serve.
That is not a technology question. It is a values question. And the answer will determine whether fintech's legacy is democratisation or a more sophisticated version of what came before.