Back to Work
Lead UI Designer

SBS Global
Design System.

A ‘Universal design system’ for Sopra Banking, a single source of truth serving multiple products and regions, built mobile-first with accessibility and RTL support.

Client
SBS products from Spain, France, Belgium and Dubai
Team
Five designers across Spain, Belgium, the UK and Dubai
Scope
Tokens, components, docs
Platforms
Responsive web and mobile
Live docs
Zeroheight
Summary

A company built by acquisition, and an interface that showed it.

Sopra Banking Software was part of Sopra Steria. It had grown by buying banking software companies in different countries, so its products looked and worked differently from each other.

The importance of great design has gained traction throughout the company, and a project was launched to create a 'Universal design system' to be used throughout Sopra Banking products. The benefits of creating such a system are:

  • Products would be easily identifiable as SBS products
  • A single source of truth for all design
  • Design could be done at speed by all our different businesses, using easily identifiable components

The Universal Component Design system project was launched in October 2020. I led the design side of it, running a team of five designers based in Spain, Belgium, the UK and Dubai. Four countries, three time zones, one eight-week build.

Methodology:

  • Double diamond
  • Agile sprints
  • Kanban structure in stand-ups

Tools used:

  • Figma
  • Zeroheight
  • Notion
  • Miro

Deliverables:

  • Design audit created from four of our products
  • Figma file containing all the components and styles
  • Zeroheight link containing the design principles, style and component documentation

The eight-week plan

Week 1-2WC 19th & 26th Oct
Week 3WC 2nd Nov
Week 4WC 9th Nov
Week 5WC 16th Nov
Week 6WC 23rd Nov
Week 7WC 30th Nov
Week 8WC 7th Dec
Mobilisation + inventory
Sprint 0Define requirements
Sprint 1Design production
Sprint 2Design production
QA + wrap up
  • Identify components used across products
  • Define lists of components to design
  • Define structure of pattern library
  • Define design principles
  • Define design direction routes
  • Establish design direction (look and feel)
  • Test designs on product screens
  • Set up pattern library environment in publishing tool
  • Set up pattern library navigation and page structures
  • Start documentation of components
  • Iterate look and feel
  • Create small set of initial components
  • Complete documentation for those components
  • Test designs on product screens
  • Iterate look and feel
  • Design a wider set of components
  • Continue documentation
  • Test designs on product screens (2 screens only)
  • Iterate look and feel
  • Design a wider set of components
  • Continue documentation
  • Test designs on product screens (2 screens only)
  • Finish design of remaining components
  • Finish documentation
  • Tidy up inconsistencies in design or documentation
  • Ensure the library is ready for handover

Governance for the pattern library was defined in parallel with the design work, not after it.

Aims

One source of truth, usable across multiple products.

The aim of this project is to create a 'Universal design system' that is a single source of truth, which can be used across multiple SBS products. Following the mobile-first approach, it has been decided that the components will be designed as responsive while taking into consideration accessibility.

A single set of styles

One product design pattern library holding a universal set of styles, rather than a library per product.

Guidelines beside it

An accompanying set of guidelines to promote consistency, so the system holds without us in the room.

Inclusive by default

Mobile-first, responsive, accessible, and working in right-to-left as well as left-to-right.

Design thinking process

Double diamond.

The Double Diamond is a design process model developed by the British Design Council in 2005. It divides work into four phases: Discover, Define, Develop and Deliver, and its defining feature is the emphasis on divergent and convergent thinking: many ideas are created, then refined and narrowed to the best one. That happens twice, once to confirm the problem definition and once to create the solution.

The value of the model on this project was permission. It gave me a shared language for why the first two weeks produced no designs, which is the hardest fortnight of a design system project to defend.

DiscoverDefineDevelopDeliverWidest point: 700+ componentsWidest point: every component, 3 platforms, 2 directions4 products, one brief5 principles, one direction21 components, published
The model with this project’s numbers on it. The two pinch points are the only places a decision was actually forced: what the problem was, and what we would ship.
Discover / Weeks 1-2

We counted what was actually there.

Timeline
Two weeks assigned to complete discovery.
What
Four SBS products were identified with key stakeholders, on which we would conduct a component audit.
How
With each product owner’s permission we went through every product and screenshotted every component we found on every page.
Result
More than 700 different components, laid out together in Miro.
Miro board showing the full component inventory
The full inventory end to end. Seeing it at this scale is what made the case for the project unarguable.

Example: the same icons across different products

Icons used across three SBS products 1Icons used across three SBS products 2
Delete, search, save and export in three of the four products. Nothing was wrong on its own; together it was unlearnable for anyone using more than one product.
Define / Week 3

Every principle answers a problem we found.

I ran workshops with key stakeholders to identify the foundations and principles our designs would be based on. Each principle exists because of a specific failure we could point at in the inventory, which made them easy to defend later when a decision was contested.

Inconsistent UX, design and copy across the portfolio
Modular
Difficult to customise and rebrand for customers
Adaptive
Does not reflect Sopra strategy and ambition
Inspirational
Early stages of building a design culture
Intuitive
Does not meet disability regulation standards
Inclusive

Design style

We built moodboards to identify which style to adopt: flat and monochrome, or shading and colour. The result was a combination of both — flat and minimal, with shading and colour used to carry meaning rather than decorate.

Two moodboard routes
The two style routes taken to stakeholders.
Decision / Define

Treat Arabic and right-to-left as a foundation, not a localisation phase.

SBS ships to Dubai. Arabic sat on the plan as a later phase, the way localisation usually does, build the system, translate it afterwards.

I moved it to the front. Mirroring is not translation. It changes component anatomy, icon direction, number and date formatting, the reading order of a table, where a calendar’s navigation controls belong, which side an error attaches to. Retrofitting that into a finished library is not an adjustment, it is a rebuild.

It cost us a little time per component. It saved the system from being unusable in one of its four markets.

The obvious move

Ship the English system, handle Arabic as a localisation workstream later.

What I did

Made RTL a condition of a component being finished. No component was signed off without its mirrored variant documented.

Develop / Weeks 4-7

Two-week sprints, reviewed daily.

Working in two-week sprints, we organised the work using Kanban in our daily stand-ups, and held weekly feedback sessions with key stakeholders to make sure we were progressing in the right direction at speed.

UX focused on each individual component, fleshing out its functionality and feeding requirements back to UI, who were fast followers in designing components against those requirements. At the end of each day the team reviewed the components designed, checking they fulfilled everything asked of them.

Decision / Develop

Rebuild the foundations before touching a single component.

The instinct in the room was to start with buttons. They are visible, they demo well, and after a 700-component audit everyone wants to see the pile shrink. Instead, I spent the first sprint on type scale, spacing, grid and colour. We could not get down to 21 components until we agreed what a component was made of. Otherwise we would only have fixed the surface of the inconsistency and left the cause in place.

Once the foundations were fixed, component work moved faster than the schedule assumed, because every decision inside a component had already been made once.

The obvious move

Start with the most-used components and show visible progress in week one.

What I did

Spent the whole first sprint on tokens, type and spacing, and explained why the visible progress comes later.

The first mock-up and style tile

First mock-up and style tile 1First mock-up and style tile 2
Presented at the end of the first sprint: a product screen in the new direction, and the style tile behind it. Type scale, colour, buttons, form fields and table headers existed as a set before any component was drawn from them.
Deliver / Week 8

What we handed over.

  • 21 fully formed, fully functioning responsive components with full CSS styling.
  • A full set of responsive grids for desktop, tablet and mobile screens.
  • Full documentation of each component and its states, with how and when to use it, alongside the foundations and design principles.

That formed the foundation of a universal design system for SBS. Components would continue to be added as new projects and products identified missing functionality, which is why the governance model mattered as much as the components did.

Decision / Deliver

Make Zeroheight the source of truth, not the Figma file.

A Figma library is a source of truth for the people who open Figma. Engineers in four countries mostly do not.

So the published documentation became the system, and Figma became one of its inputs. Each component was documented with usage guidance, code, native behaviour and its right-to-left variant on separate tabs of the same page — so a developer, a designer and a product owner could all arrive at the same place and each find the answer they came for.

Governance was defined in parallel with the design work rather than after it, so a route for contribution existed before anyone needed one.

The obvious move

Hand over a well-organised Figma library and a set of slides.

What I did

Published to Zeroheight as the canonical reference, structured around who is reading rather than how the file is built. It is still live.

Zeroheight component overview
The finished library in Zeroheight: twenty-one components, from accordion to toggle.
Footer component variants
One component in full. The footer at large and small breakpoints, with language switcher, links, and combined variants — every state specified rather than left to interpretation.

Example of one product screen

Product screen before and after 1Product screen before and after 2
The original product design, and the same screen rebuilt from the system. The same information, the same density, roughly the same screen count. What changed is that a banker moving between SBS products now meets the same conventions in each one.
Twenty-one components covered every screen we tested them against.
4
products audited, across four countries
700+
distinct components found in the inventory
21
components it was eventually reduced to
3
platforms covered: web, iOS and Android
Applying it / Fidor

Onboarding redesign.

A system is only true when someone else builds with it. We reskinned Fidor’s onboarding flow against the design system components, on mobile and desktop, using the real flow rather than a sample screen.

Fidor onboarding before and after
Original design above, reskin following our design system components below, for both mobile and desktop.

Admin portal redesign

Fidor admin portal before and after
The Fidor admin portal, before and after. Dense, table-heavy screens are where a component library either holds or falls apart.
Applying it / SAB

SAB redesign.

A second acquired product, running on the same components: marketing pages, login, transfers, account summaries. Different content, different density, same system.

SAB before and after
Original designs on the left, reskins using the design system on the right.
Mobile native components

Android and iOS, matched to the system.

Each component was specified for web, iOS and Android. Where a platform would not bend, we documented the constraint rather than pretending it away — iOS holds San Francisco on a back button, Android fixes its date picker font, and a system that ignores that produces specs nobody can build.

Date input across web, iOS and Android
Notification behaviour
Right-to-left

Arabic components, documented in Zeroheight.

This is what the RTL decision produced. Every component carries a right-to-left tab alongside usage, code and native — not a note saying Arabic is coming, but the mirrored component with its own anatomy, states, variants and error handling.

Zeroheight page showing usage, code, native and RTL tabs
One component, four tabs. The right-to-left tab sits beside usage, code and native rather than in a separate document.
Arabic anatomy and interactive states 1Arabic anatomy and interactive states 2
Anatomy and interactive states in Arabic. Labels, field, calendar icon, month and year controls, and day — each mirrored and numbered as they are in the English version.
Arabic variants and error handling
Variants and error handling, including the mixed case: an Arabic interface where numerals are required in English.