The Investor Lens: Why Trust-First Lending Is the Better Bet

Lending, Reimagined, Part 7 of 7

A lending business can report a strong month while its customers have a difficult one. Disbursals rise. Revenue rises. More people borrow again. The dashboard looks healthy.

The questions I would want an investor to ask are what those customers used the money for, whether the commitment remained manageable, and why they came back.

Repeat borrowing can reflect a useful ongoing relationship. It can also reflect a shortfall that the previous loan did not solve. The revenue line will not explain the difference on its own.

Across this series, we have examined expensive credit, the borrower’s experience, fintech incentives and the role of regulation. The final question is commercial: can a company put suitability first and still become a valuable business?

I believe it can. But trust is a hypothesis to prove through behaviour and economics. Calling a company user-first is not evidence that the model works.

Start with what the company is being paid to do

A lender, a loan distributor and a personal finance product have different economics. They may share an interface, but they do not carry the same costs or risks.

A lender has to account for funding, credit losses and capital. A distributor must understand acquisition costs, partner payouts, servicing obligations and any risk-sharing commitments it has actually accepted. A broader financial assistant has to earn continued use and pay for the technology and support that make it useful.

An investor should be able to trace a rupee of revenue back to the event that generated it. Did a customer pay for an ongoing service? Did a partner pay after disbursal? Is the amount contingent on the customer keeping a product? What happens if the transaction is reversed?

That clarity makes conflicts visible. If the company is paid when a loan is taken, recommending a smaller loan or no loan has an immediate commercial cost. A credible plan explains how the company can afford that choice.

Lifetime value should mean a useful relationship

There is an attractive version of the trust argument: help someone today, and they will return when they need you again. They may also recommend you to others.

It is plausible. It is not automatic. People can appreciate good advice without becoming frequent users, paying subscribers or valuable referrals. Some questions only arise once every few years.

The business therefore needs a reason to be useful between transactions. For Finanjo, this is where Jo belongs in the discussion.

Someone may need help understanding a rise in spending this month, planning around an uneven payday next month and comparing a borrowing decision later. Our ambition is for Jo to help with that broader sequence of financial questions, within the limits of the information available and the product’s actual capabilities.

The unit of value should be a question resolved or a task completed. A long chat, a colourful chart or daily app opening is not enough evidence. We need to know whether the person understood the answer and could use it.

That gives us a more demanding retention question: do people come back because the service helps them manage their money, including when they do not need another financial product?

Better matching has to show up after the application

A marketplace can improve conversion by routing applicants to more suitable partners. That can create value for the borrower and the lender. But an approval only tells us that an application passed a decision process.

For a debt-consolidation case, the next questions are whether the old balances were cleared, whether the total cost fell after fees and whether the new payments remained manageable. For a short-term cash-flow problem, we should ask whether the gap closed or simply moved into the following month.

These outcomes take longer to observe than a disbursal. They also require careful attribution. Customers who choose advice may differ from those who do not. A company serving less risky borrowers might report better repayment without having improved the advice at all.

I would want comparisons between reasonably similar customer groups, clearly stated observation periods and visible missing data. A small early sample can generate a useful question. It should not be presented as proof of a permanent advantage.

The economics need to include the cost of saying no

Suppose an advisor spends time helping someone decide not to borrow. That consultation has consumed real resources even if it generated no commission.

AI may reduce parts of the work, but it introduces its own costs: model usage, evaluation, engineering, data access, support and the people needed when the system cannot safely resolve a question. These belong in the model from the beginning.

For an investor, the useful figure is contribution after the relevant variable costs and obligations, alongside the continuing fixed costs required to operate the service. Gross partner revenue alone can make a model look healthier than it is.

The same discipline applies to acquisition. Referrals can be valuable, but they are not necessarily free. Incentives, support and attribution all matter. We should distinguish customers acquired through genuine recommendation from customers responding to a reward.

The thesis is that a useful relationship can eventually support more efficient acquisition and appropriate monetisation. We should test that thesis in cohorts over time rather than inserting an optimistic lifetime-value estimate into a pitch deck.

Distribution success does not remove the conflict

Credit Karma is a useful example because its model is easy to describe: free financial information and recommendations, with commissions typically earned when recommended products are approved. The company explains the model here.

It is also a reminder that the presentation of recommendations matters. In January 2023, the FTC finalised an order requiring Credit Karma to pay $3 million to settle allegations involving misleading pre-approved and approval-odds claims. Read the FTC’s account.

The lesson is not that commissions make useful guidance impossible. The lesson is that the conflict needs controls that survive pressure to improve conversion.

Who can change the ranking logic? Can a commercial team buy a better position for a partner? Are approval estimates presented with appropriate uncertainty? Are complaints reviewed independently of the team responsible for sales?

An investor should ask to see those controls in practice. A founder’s intentions are a starting point, not an operating system.

A scorecard that would change the conversation

I would want a board review to put commercial performance next to evidence of customer value. Neither should be used to excuse the absence of the other.

• Acquisition: cost by channel, qualified activation and the proportion of users who complete a useful first task.
• Customer outcomes: understanding of costs, completion of the intended debt replacement, repayment difficulty and complaints, with the limits of the evidence stated.
• Retention: return to useful financial tasks, including activity that does not involve taking another loan.
• Economics: revenue and contribution by cohort after relevant servicing costs, incentives, reversals and risk obligations.
• Independence: recommendations changed after review, cases ending without a sale, partner concentration and the effect of payout changes on rankings.

A high rate of no-sale decisions is not automatically good. It might reflect careful advice, poor matching or a product that fails to help suitable borrowers. The point of the measure is to investigate the decision, not to reward rejection.

Likewise, more time in the app is not automatically good. If Jo answers a question correctly in a minute, forcing a longer session would make the experience worse.

What could make the advantage durable?

Access to a model or a connected account is unlikely to be enough on its own. Other companies can adopt similar tools and build attractive interfaces.

The harder work is a reliable process: understanding the question, obtaining appropriate information, recognising gaps, explaining the answer, helping the user act when authorised and following up when it is useful.

Over time, that process may produce valuable learning about which explanations and interventions help. But financial data is not a resource to collect without limits. Permission to read information for one purpose is not permission to use it for every other purpose, and access to data is distinct from authority to transact.

The durable advantage we should aim for is earned reliance. A person returns because previous answers were useful, limitations were clear and support was available when needed. We have to keep earning that reliance with each interaction.

The bet we want to make

Trust-first lending is the better bet I want Finanjo to test. It asks more of the company: a useful product outside the moment of borrowing, visible conflict controls, honest unit economics and evidence of customer outcomes.

It also asks more of investors. If a board rewards only disbursals, founders will feel that pressure. If it examines how customers fare after borrowing and whether a broader financial relationship is developing, it gives the company room to build for a different result.

This series began with the cost of credit. It ends with the kind of company we want to build around people’s money.

Loan Buddy and Jo should help someone reach a decision they understand, including a decision that earns us nothing today. Our job is to make that useful enough to sustain a business. The evidence will have to follow the ambition.

This essay sets out a business thesis and proposed measures. It does not claim that Finanjo has already demonstrated these financial or customer outcomes. References were checked on 6 October 2026.

Previous: The Regulator’s Dilemma: What Policy Could Do Next
Read more from the Founders Office