At some point in almost every fintech product conversation, someone says a version of the same thing:

"We need this feature to drive revenue."

It's a legitimate goal. Products need to be commercially viable. Features need to justify their development cost. Nobody builds a financial product for altruistic reasons alone.

But in my experience, the moment a data-driven feature is designed primarily to sell — before it has earned the right to — is the moment it starts failing. Not loudly. Quietly. Users ignore it. Disable it. Route around it. And eventually stop engaging with the product that hosts it.

There's a sequencing principle I've come to apply to every data-driven feature I work on. It goes like this: accurate first, helpful second, sell last. It sounds obvious. It is almost never followed.

First: be accurate

Before a feature can do anything useful, it must be trustworthy.

In financial products, accuracy isn't just a quality metric. It's the foundation of the entire user relationship. When someone opens a banking app and sees an insight about their spending, their income, their financial health — they're extending a form of trust. They're treating your output as a reliable signal about their own life.

An insight that is sometimes wrong — even occasionally — doesn't just create a bad user experience. It erodes the credibility of every insight that follows. Users who are told something incorrect about their money don't forget. They discount everything else your product tells them. The feature might still exist on the screen, but it's already dead.

This is especially true in thin data environments, where the temptation is to surface insights before the underlying data is robust enough to support them. The pressure to show something — anything — to stakeholders and users is real. Resisting that pressure, and waiting until the accuracy is genuinely there, is one of the hardest and most important product decisions you'll make.

Second: be helpful

Accuracy is necessary but not sufficient. A feature can be completely accurate and still be useless.

There's a meaningful difference between a feature that shows someone their spending and a feature that helps them understand what to do about it. Between a dashboard that displays numbers and one that surfaces the number that actually matters right now. Between an alert that fires correctly and one that fires at the moment it can actually change a behaviour.

Helpfulness in financial products is about relevance and timing. It's about surfacing the right insight at the moment when a user can act on it — not just when the data is available. A correct insight delivered at the wrong moment is noise. A correct insight delivered when the user is making a decision is value.

Getting to helpfulness requires knowing your users well — how they actually behave, not how you assume they do. It requires building feedback loops into the feature so you can learn whether it's changing behaviour over time. It requires being honest with yourself when a feature is technically working but not actually helping anyone.

Last: sell

Revenue, conversion, upsell — these are legitimate business goals. They are also the natural outcome of features that have genuinely earned accuracy and helpfulness first.

The financial features that drive the most sustainable commercial outcomes are the ones users trust and rely on. When a feature has demonstrated — through consistent accuracy over time — that it understands a user's financial life, users become receptive to its recommendations. A suggestion to open a savings account from an app that has spent six months giving you reliable spending insights lands very differently than the same suggestion from an app you've never trusted.

Rushing to sell before accuracy and helpfulness are in place doesn't accelerate revenue. It undermines the credibility that would have made the revenue possible.

Business teams will not love this sequencing. Leadership will push for commercial outcomes faster. The pressure to demonstrate ROI on a feature before it has had time to earn user trust is real, constant, and understandable.

But skipping steps one and two to reach step three is the fastest route to a feature that users abandon — taking the commercial opportunity with it.

Why this is really about trust

Underneath the sequencing principle is a simpler idea: trust is infrastructure.

In physical banking, trust was built through decades of presence — the branch on the high street, the institution that had been there longer than you'd been alive. Digital banks don't have that. They have to earn trust through the quality of their product interactions, one feature at a time.

Every accurate insight is a deposit into a trust account. Every helpful recommendation is another deposit. And when the commercial moment comes — when the product suggests something that would benefit both the user and the business — there's enough trust in the account to make it land.

The reverse is also true. Every premature insight, every inaccurate recommendation, every feature that prioritised the sale over the user — these are withdrawals. And in a trust account with a young balance, you can go negative very quickly.

Building for accuracy first isn't slow. It's the only approach that actually compounds.