All work

01Sydbank · 2024

Enhancing the UX of data governance

A Master's research project reimagining how a bank shares and trusts its data — where I led the prototype and usability testing.

UX ResearchData GovernanceFigmaPrototyping
Sydbank data catalog — New way to work with data
Partner
Sydbank
Context
MSc thesis · SDU
My focus
Prototype & usability test
Method
Mixed-methods · 5 studies

Introduction

Sydbank's Alation catalog stores thousands of tables, queries and articles — yet employees still solve everyday data questions over Teams, email and hallway chats. That siloed habit buries good work and slows new analysts. In a team of four at SDU, I led Study 3: the high-fidelity prototype and the usability test with real Alation users.

Research question

How might we improve Alation's community features so the catalog actively drives knowledge-sharing between its users?

Feature overview — new layout, threads, tags, ratings and navigation
System overview — the acknowledged UX patterns we built on.

Research method

Five studies, each feeding the next.

A mixed-methods approach where each study built on the last — so decisions stayed grounded in real behaviour, not assumptions.

  1. Study 01

    Interview analysis

    Re-analysed Sydbank's internal interviews: ratings, conversations and queries were useful in principle but overlooked in practice.

  2. Study 02

    Observation & heuristic evaluation

    Netnographic review plus an expert pass against Nielsen's 10 heuristics — confirming near-zero community activity and the UI issues behind it.

  3. Study 03 — my focus

    Prototype & usability test

    Built the high-fidelity Figma prototype — Threads, a redesigned dashboard, clearer states — and tested it unmoderated in Maze.

  4. Study 04

    Mixed-methods data

    Triangulated Maze behaviour (heat-maps, completion, time-on-task) with qualitative email responses, coded in NVivo.

  5. Study 05

    Communities of Practice lens

    Framed it all through Wenger's Communities of Practice — UI alone can't manufacture a culture of sharing.

Results

9
Test participants
8/9
Completed all tasks
5.5/6
Top task usability rating
5
Linked research studies

Key features

Collaboration, woven into the catalog.

A high-fidelity prototype layering community features onto familiar workflows.

  1. 01

    Threads

    A tag-driven discussion anchored to every table, query and article. Questions get asked in context and answers marked “Accepted” — so the next person finds a trusted resolution.

  2. 02

    Curated home dashboard

    Latest Threads, Popular Queries and Featured Articles surface active content at login, replacing the static landing page with a living overview.

  3. 03

    Clarified rating UI

    Ambiguous flags become clear positive / negative thumbs with colour-safe states and a tooltip explaining each rating's impact.

  4. 04

    Undo toasts & micro-animation

    Real-time feedback makes actions reversible and the UI feel tactile — WCAG-AA throughout.

Grounded in

Decisions backed by principle.

  • Nielsen's 10 heuristicsAudited and redesigned the UI against each principle.
  • Fitts's LawSized interactive targets for faster, surer hits.
  • Jakob's LawMatched mental models from LinkedIn & Gmail.
  • WCAG 2.1 AAContrast-safe states across every component.
  • What do prototypes prototype?Houde & Hill framing to pick the right fidelity.
Curated home dashboard with Latest Threads, Queries and Articles
The curated home dashboard surfaces active content at login.
Threads discussion panel anchored to a data asset
Threads turn one-off questions into reusable, accepted answers.
Compose modal for posting a thread with query, code and mentions
Composing a thread — inline queries, code, images and @mentions.
It's not just ‘empty’ when I open it — I can see my colleagues are active, and I keep up with what Alation can do.
Sydbank employee · usability test

Reflection

The honest read: usability improved, but a single prototype can't create a community of practice on its own.

  • 01Half of participants didn't expect to use the features more — the gap is partly workflow and culture, not just interface.
  • 02Non-anonymous responses risked social-desirability bias; I'd anonymise participation next time.
  • 03With more time, I'd run a moderated before/after test for step-level insight.