Spending Insights

Defined the zero-setup category system and 38-feature roadmap for a money-management feature now live for millions of customers across Lloyds, Halifax and Bank of Scotland.

The Problem

Customers couldn’t see where their money was going without exporting statements into a spreadsheet. Category-based budgeting tools existed across the market, but most demanded upfront setup — mapping accounts, naming categories, tagging transactions — before they delivered any value, and most people never finished onboarding. For a bank serving a huge, varied customer base across three brands, a configurable tool wasn’t going to reach the people who needed it most.

The brief: give every customer a clear, zero-effort picture of their spending from day one.

My role / the team

I owned the category system and feature definition. Worked closely with a product owner, a researcher, and platform engineering leads across three brand teams. The team delivered a 38-feature roadmap and shipped the first release to millions of customers.

What The Evidence Said

Usability sessions on early concepts showed customers abandoned any flow that asked them to configure categories before seeing value. Diary studies found people checked spending in short, frequent bursts — a few seconds at a time — not in long analytical sessions. And a review of support contacts showed miscategorised transactions were the single biggest driver of trust loss in existing PFM tools, more damaging than having too few categories.

The Judgment Calls

Eight categories, not the seven research suggested.

Research pointed to seven clean categories. I added an eighth — an honest ‘Everything else’ bucket — because forcing every transaction into a tidy category was exactly the kind of overreach that had eroded trust in competitor tools. Better to admit the system’s limits than pretend precision it couldn’t deliver.

Zero-input by default, over a configurable setup flow.

Configurability tests well in a boardroom and badly in the real world — most people never finish setup. I traded flexibility for immediacy: categorisation happens automatically, with no onboarding flow. The cost was less personalisation for the small number of customers who wanted to fine-tune everything; the trade-off was worth it at this scale.

A deliberately thin Day 1, against a fully-featured wishlist.

Stakeholders wanted budgets, alerts and forecasting in the first release. I scoped Day 1 down to categorisation and a monthly summary only — the smallest thing that could be trusted — and sequenced the rest across a 38-feature roadmap. Shipping a narrow, reliable v1 mattered more than shipping a broad, shaky one.

The Design

Drop the SpendingInsightsMotion component here

The pilot, animated: January’s quiet month against December’s Christmas spike — same eight categories, zero setup.

The category summary — the single screen most customers will ever see.

The honest ‘Everything else’ bucket, shown rather than hidden.

Monthly summary — the only Day 1 comparison view.

December’s Christmas spike, shown against January’s quiet month.

What Shipped & What Happened

Day 1 shipped to all three brands and is now live for millions of customers. Engagement has been strong enough to justify continued investment against the 38-feature roadmap, with the next releases focused on trends and light forecasting. Support contact volume related to categorisation stayed flat rather than spiking, which — given the honesty-first approach to the ‘Everything else’ bucket — we read as a quiet win.

What I’d Do Differently

I’d bring platform engineering into the category-taxonomy decisions earlier — some of the ‘Everything else’ edge cases were a data-mapping problem as much as a design one, and we spent time relitigating decisions that better upfront alignment would have avoided.