One language for
a £3B business

The strategic problem

Different colour schemes, chart types and hierarchies on every dashboard. The same metric labelled differently — or calculated differently — depending on who built it. Filters, date ranges and drill-downs each handled their own way. And when none of it worked, screenshots and manual copying.

The cost was not aesthetic. Leaders could not compare performance across business units without first mentally translating each dashboard’s logic.

My role

I led the design effort across Finance, HR and Broking with a project manager and a product lead: running discovery into how dashboards were really used, defining the visual language and pattern library across two very different platforms, managing a small team, and translating hard technical constraints into UX patterns.

A good part of the job was navigating scepticism from teams who were protective of what they already had.

Realistic mockups over beautiful ones

Dashboard projects here had a mixed record. Finance and HR had seen initiatives that looked good in Figma and fell apart in implementation, so I earned trust by showing I understood the constraints.

No rounded corners in the Power BI designs, because the platform does not support them. System fonts — Arial, Georgia, Courier — not custom typography that would never ship. Mockups built on real data scenarios and actual edge cases. And fast turnaround, with feedback incorporated between sessions.

It was not about flashy design. It was about demonstrating I could deliver inside the real constraints.

A unified visual language

I defined a cross-tool pattern library covering all twenty dashboards: a categorical and quantitative palette that worked within Power BI’s limits, typography and spacing tuned for reading numbers at a glance, chart-type guidance with the rationale attached, and standard patterns for filters, tabs, tables and drill-downs.

The test was simple. Someone who understands one dashboard should feel at home in any other, whichever tool it was built in.

Standardised filtering, and a glossary

Filtering was the single biggest source of confusion — where filters lived, how they behaved, whether you could tell which were active. I put a stake in the ground: consistent global-versus-local placement, standardised multi-select, reset and defaults, clear surfacing of active filters, and one date-range pattern for time comparisons.

Then the less visible half: how to represent negative numbers (brackets, not minus signs), currency formatting, metric definitions with their calculations, cohort versus total labelling.

None of this is glamorous. But for a data product, trust is the core UX feature.

Export UX

Discovery made it clear that the dashboard was rarely the destination. Leaders took data out — into Excel to analyse, into PowerPoint for the board, into email to circulate — and the existing export was awkward, dropped filters and gave no control over what came with it.

So I designed one export flow: column selection, filter preservation so the export matches what you are looking at, date range and aggregation controls, and formats matched to what happens next.

It got none of the attention a visual redesign gets. It directly addressed how executives actually used the data.