Royal London · Workplace Futures

Getting a clear answer when the research said “it depends”

Royal London is building a new online platform for employers managing workplace pensions. Two decisions from my first months on it: how its data tables should filter, and how to make its pages consistent without waiting on anyone.

Role
Senior UI Designer
Dates
Mar 2026 to present
Product
Workplace Futures, a workplace pensions platform for employers
Team
UI specialist in a mostly UX team, with a user researcher, accessibility specialists and engineers
Scope
UI library used by ~10 designers, referenced by ~20 engineers
Status
In progress
Column-header filters (A) and a filter drawer (B), the two patterns we tested. Recreated for this case study; details changed.

The problem

Workplace Futures replaces many of Royal London's existing services for business customers. One of the teams I support needed tables for pension contributions and related data.

The people using these tables are business managers and finance operators. They work through a lot of data and need to narrow it down quickly and accurately.

We had two ways to filter: Excel-style filters in each column header, or a drawer where you set every filter and apply them together. Whichever we chose would be repeated across many tables, so a wrong call would spread across the product.

Constraints

  • Financial data. Accuracy and clarity matter more than speed of interaction.
  • Accessibility. Both patterns had to work by keyboard and with a screen reader. I worked through this with our accessibility specialists.
  • A shared design system. The company design system serves the whole business, so changes to it take time.

My role

I designed both prototypes and wrote the test requirements with our user researcher. The researcher ran and moderated the sessions, analysed the results and presented the findings. We made the decision as a team, based on those findings.


Round one: a 50/50 split

In five moderated sessions, participants used each pattern to filter by specific fields and categories. Afterwards they chose which they preferred against set criteria: which was easier to use, and which showed active filters more clearly. We also looked at discoverability, usability and how well people completed the tasks.

The result was close to an even split, with complaints about both patterns.

  1. Round one5 sessionsStandard Figma prototypes. Close to a 50/50 split, complaints about both.Inconclusive
  2. Change of methodPrototypes that behaveBoth patterns rebuilt in Figma Make. Filters apply to real data and results update.
  3. Round two25 sessionsSame tasks, same criteria.Drawer clearly preferred
From an inconclusive first round to a clear result.

Changing the method, not the question

Our read was that the prototypes were the problem, not the patterns. Standard Figma prototypes can only follow the paths you wire up in advance. Participants were reacting to the limits of the simulation as much as to the designs.

So I rebuilt both patterns as fully interactive prototypes in Figma Make. They behaved like real software: filters applied to the data and the results updated. We then ran 25 sessions with the same tasks and criteria.

The result

The drawer, where you set all your filters and apply them at once, was the clear preference. The programme now has a tested pattern it can reuse, rather than a judgement call each team makes again.


Raising UI quality without waiting

A second problem ran alongside the filter work. The company design system serves the whole business, and a single programme moves faster than a shared system can. Pages across Workplace Futures were drifting apart.

I built a UI library on top of the company design system, scoped to the programme:

  • Plain-named tokens. Technical names like "primary 600" became "primary", and "neutral 900" became "foreground primary". Designers pick the right one without a lookup.
  • Opinionated components. Around 10 components with fewer options than the originals. The company container had several style variations; ours has one, with rounded corners and a stroke, so everyone gets the same result.
The programme library sits between the company design system and product teams.
Technical token names and the plain names that replaced them
Technical namePlain name
primary 600primary
neutral 900foreground primary
neutral 600foreground subtle
Plain names describe what a token is for, not where it sits on a scale.

To make the case, I took three existing pages that didn't match each other and redesigned them against the new guidelines. Seeing their own work side by side, teams took the guidelines up. I didn't need authority to ask for consistency; the comparison did the asking.

Three pages before and after, as wireframes. Recreated for this case study; details changed.

Outcome

  • A tested filtering pattern the programme can reuse across its tables.
  • A UI library used by around 10 designers and referenced by around 20 engineers.
  • UI now joins teams at the prototyping stage, working with UX designers and accessibility specialists before build. This is a first implementation, not yet written into the programme's ways of working.

What I'd do differently

When the question is about how something behaves, test with prototypes that behave. Next time I'd start with interactive prototypes and save a round.

Next, I want to feed what we've learned from the programme library back into the company design system.