Senior UX Designer · Design Systems

Richard Andhika

My favourite design decision is usually the one that removes something.

Utrecht, NL
About

10 years in product design and UX. Built and scaled design systems at Aegon NL (5M+ customers), unifying 60+ products under one shared component library.

Indonesian by birth, Dutch because worstenbroodjes.

Outside work
Skiing. Alps when budget allows, dry slopes when not
Emerging tech. Currently building things with AI I didn't know I needed
Photography. Animals, forests, things I don't want to forget
Expertise
Product designWCAG 2.2 AAToken architectureHTML / CSSComponent API designSystems thinkingAdoption strategyStakeholder mgmtAI prototypingData analysis (GA360, Hotjar)Saying no to feature requests, politely
Contents

Three case studies.

Case Study 01
Restructuring AIP
Aegon Intermediair Portaal. 22,000 advisors, WFT-scoped navigation, persistent search, and role-aware dashboards.
View case study →
Case Study 02
Building the System
Aeon Design System. 60+ products unified under one shared component library. Audit, conditional logic, documentation, adoption pipeline.
View case study →
Case Study 03
Driving Adoption
Hypotheek Data Table. The largest revenue team refused the system. Composable table, clear ownership split, zero accessibility regressions.
View case study →
Case Study · Enterprise B2B · Aegon

22,000 advisors.
One interface for all of them.

Multi-BU portal redesign for independent insurance advisors.

Lead Designer AIP Scrum of 8 · 7 BUs 18-month deadline
AIP 3.0 · Aegon Intermediair Portaal
Problem

Built for everyone. Designed for no one.

AIP exposed every business unit to every advisor, regardless of WFT license(!)WFT is the Dutch financial advisory qualification. It determines which product categories an advisor is legally allowed to advise on., active portfolio, or daily workflow.

"Ik doe alleen Leven. Maar hier staat ook gewoon Schade, Pensioen, alles."
"Er staat zoveel in dat menu. Ik weet eigenlijk niet eens wat de helft is."
"Ik open het en zie meteen tien dingen die ik nooit nodig heb. Alles staat door elkaar."
Task Restructure AIP so advisors could get to the right client, the right product, the right task faster.
AIP 3.0 homepage with role-scoped tiles, role-aware dashboard, and persistent search
AIP 2.0 homepage with 7-BU mega-nav and news feed
drag to reveal
AIP · Before & After
From 7-BU mega-nav to role-scoped workspace.
Left: AIP 2.0 · every business unit, every link, surfaced to every advisor. Right: AIP 3.0 · only the BUs this advisor is licensed for, with persistent search and a personalized landing.
Evidence

The portal was role-blind.

Three months of analytics, session recordings and advisor interviews all pointed at the navigation structure.

01
Behavioral analytics
GA360 · 12-month window · all advisors
02
Session recordings
Hotjar · 200+ tagged sessions by BUs
03
Advisor interviews
50+ advisors
Synthesis insight
"Different advisor profiles were running on one undifferentiated portal. Every link visible to everyone, every session."
GA360 nav-link engagement chart
GA360 nav engagement
By the numbers

One portal. Three ways it failed advisors.

Analytics, heatmaps and the portal audit showed the same pattern.

85%
of interactions landed on 5% of nav links
GA360
3
clicks to search a new client mid-session
Heatmap sessions
7
BUs visible to every advisor, regardless of role
Portal audit
20+
tasks handled per advisor, on average
GA360
Approach

Three design decisions, mapped to three advisor pain points.

Persistent customer search
Recent bekeken
Laatste activiteit
Always one input away.
addresses3 clickssearch depth
WFT-scoped navigation
3 of 7 · licensed
Only the BUs this advisor is licensed for.
addresses7 BUsno role scope
Role-aware dashboard
Klant
Acties
Pending actions, portfolio, BU metrics.
addresses85% / 5%nav imbalance
AIP 3.0 persistent search visible in context
Persistent search
Search from any page.
A persistent field in the top bar, always visible, with recent searches and recent clients.
AIP 3.0 WFT-scoped navigation comparison showing all business units versus advisor-licensed business units
WFT-scoped navigation
Only what this advisor is licensed for.
Hiding an unlicensed business unit is good UX and good compliance.
AIP 3.0 value stream dashboard detail view
Role-aware dashboard
Decisions

What shipped. What I killed in testing.

● Shipped 3 decisions
Persistent customer search
Always visible in the top bar, with recent searches and recent clients.
WFT-scoped navigation
Only the business units this advisor is licensed for.
Role-aware dashboard
Per-role task context with pending actions, active portfolio, and BU metrics.
✕ Killed in testing 1 concept · before engineering
Full-page customer search concept that was killed in usability testing
Loved internally. Wrong in use.
The full-page search concept destroyed nav context in testing. Advisors lost their place mid-task.
Delivery

18-month deadline. Four validation gates.

Each phase shipped only after the previous one validated against advisors and stakeholders.

01Jul – Sep
02Oct – Dec
03Jan – Mar
04Apr – Jun
Live
Design
Research synthesis · structural reframe · BU alignment
Persistent search · full-page search bar concept
Role-scoped dashboard · WFT nav
Cross-platform content flow · in-portal story pages
Phased rollout · 22,000 advisors
Validate
Stakeholder sign-off across 7 BUs
Gate passed
Search proven in testing
Full-page bar
Gate passed
Tested with advisors · WFT scoped
Gate passed
Journey moved to aegon.nl
Portal content
Gate passed
90%+ DS adoption · 6 months early
Shipped
Outcomes
Shipped in 12 months.
Load time down 46%.
Advisors on task up 60%.
Sources
Avg page load, 60-day window post-launch · GA360 advisor session data, pre/post · DS adoption tracked vs 18-month deadline
Reflection

What worked. What I noticed too late.

What worked
What I noticed too late
The reframe was the project
Once we called it role-blind, the three decisions were obvious. Without it, we'd have shipped a tidier version of the same problem.
I trusted GA360 before I trusted advisors
The insight was in the interviews first. I weighted the quantitative read too early. Next time the qualitative gets equal weight from day one.
DS reuse saved six months.
90%+ of components came from Aeon. That's how an 18-month plan shipped in 12.
Role taxonomy was too clean
WFT licenses gave us a defensible scope. They also flattened assistants, back-office and team leads into edge cases they weren't. Around 15% of sessions.
Case Study 02 · Design Systems · Aegon

Aeon Design System.
60+ products, one system.

Shared component architecture across Aegon NL products.

Design System Designer Aegon NL · 5M+ customers 2 senior UX designers + 3 engineers + PO
Aeon Design System
Problem

Every team had a button. None of them had the same button.

Each platform built independently. No shared components, no token language, no documentation. Every team solving the same problems in isolation.

Role · What I owned
Component audit + architecture
Auto-layout + conditional logic
Stakeholder alignment
Component brief template
Adoption strategy
Documentation guidelines
Semantic color tokens
Old platform screenshot 01
Old platform screenshot 02
Old platform screenshot 03
Old platform screenshot 04
Old platform screenshot 05
Evidence
AIP 3.0 after the redesign: role-scoped navigation with persistent search and a personalised advisor landing page.
AIP 2.0 before the redesign: a six-business-unit mega-navigation exposing every link to every advisor.
drag to reveal
AIP · Before & After
The same portal. Four years apart.
Left: AIP 2.0, each business unit shipping its own patterns. Right: AIP 3.0, one token system, one component set, 22,000 advisors on a consistent interface.
Overview of the Aeon design system: colour, typography, and the shared component library spanning 60+ products.
Aeon Design System
From 47 variants to one shared language.
Buttons, fields, date pickers, step indicators. Every component documented with states and usage rules. Built for three platforms.
Buttons
Hypotheek Portaal
Click to reveal Aeon
Aeon · Buttons
Text Input
Beleggingsplatform
Click to reveal Aeon
Aeon · Text Input
Naam
Postcode
IBAN
Dropdown
Life & Pensions
Click to reveal Aeon
Aeon · Dropdown
Badges
Pensioenportaal
Click to reveal Aeon
Aeon · Badges
Status badges
Primair Secundair Fout Geldig Waarschuwing
Filter pills
Toggles
Zorgverzekering
Click to reveal Aeon
Aeon · Toggles
Aanvullende vergoedingen
Betalingsfrequentie
Architecture

Three platforms. One component.

Platform-specific Field variants were replaced with a single component driven by context flags. Reduced build surface, eliminated drift, and cut load times by 25%.

Input component properties
Alert and notifications component properties
File upload component properties
Documentation & Adoption

Docs nobody read. Then docs they pointed to.

Scattered, inconsistent, rarely referenced. I restructured the system documentation: usage guidelines, do/don't examples, token mapping, team contribution notes.

89% of adopting teams pointed to documentation as the reason they could switch without slowing down.

Aeon documentation overview
Documentation overview screenshot
Documentation detail screenshot
Outcomes
Form completion up 25%.
Dev time down 40%.
Component adoption above 90%.
Sources
GA360, baseline vs post-rollout · Dev lead reports across 3 teams · Tracked screens, 3 teams, rollout phase 1
Case Study 03 · Design System Adoption · Aegon

The Data Table they wouldn’t adopt.
Shared structure, local complexity.

Turning a resistant revenue team into design-system partners without flattening their domain complexity.

Design System Designer Aegon NL · 5M+ customers 2 senior UX designers + 3 engineers + PO
Hypotheek Data Table
Problem

Largest revenue team. Openly resistant.

"Our tables are too complex for your system."
"We can't wait weeks for approvals."
"We'll miss deadlines if we have to conform."
Three different custom data table interfaces side by side, each built separately by a different product team.
Insight
"I'm not asking you to simplify mortgages. I want to simplify everything that isn't mortgage-specific."
Mortgage review screen showing the data table in its real product context.
Evidence

70–80% of what they needed already existed.

Shared fundamentals
Sorting, pagination, selection, bulk actions, accessibility, keyboard behavior
Mortgage-specific complexity
Conditional columns, inline calculations, warning states, expandable rows
Feature audit table separating shared data-table fundamentals from mortgage-specific edge cases, with ownership assigned per row to the design system, the platform, or the Hypotheek team.
Approach

Clear ownership. Neither team steps on the other's toes.

Hypotheek
layer
Business logicDomain rulesCustom renderingConditional columnsExtension contentUser actions
API
boundary
renderCell()renderExpandedRow()isVisible()
Design
System
UI infrastructureInteraction patternsVisual designAccessibilityPerformance
Solution

Shared core, team-owned cells.

I designed the composable model and its components. Structure, behaviour, and accessibility shared; cell rendering, expansion slots, and conditional columns owned by Hypotheek.

Their edge cases stayed. No fork required.

①–⑦ DS infrastructure·⑧ Team extension slot
Anatomy diagram of the shared data table component, labelling header, row, cell, and control regions.
Data table interaction states: default, hover, selected, loading, empty, and error.
Mortgages-in-application screen rebuilt on the Aeon data table component.
The original mortgages-in-application screen before the Aeon data table was adopted.
drag to reveal
Hypotheek · Before & After
The same flow. Before and after Aeon.
Left: the legacy in-aanvraag UI, custom-built and detached from the system.
Right: rebuilt on Aeon components.
Mortgages-in-application screen with an expandable row opened to reveal mortgage detail.
Hypotheek in Aanvraag · Aeon
Domain-specific row expansion slots.
Custom cell renderers, custom mortgage-specific logic. The spacing, animation, and behavior are handled by the design system.
Responsive

The same contract. On a 390-pixel screen.

The extension points travel. Mortgage logic keeps rendering its own cells; the design system keeps owning sorting, selection, and keyboard behaviour.

Mortgage applications list on mobile, showing filter chips and two-line client rows.
An expanded row on mobile revealing loan amount, interest rate, end date, and monthly cost inline.
Client dossier sheet on mobile with status, loan amount, and a timeline of application events.
Client documents screen rebuilt on the Aeon data table component.
The original client documents screen before the Aeon data table was adopted.
drag to reveal
Klant Documenten · Before & After
Delivery

Their biggest fear was waiting on us.

01
Direct weekly sync
Every open question gets a live answer that week. No ticket, no queue, no waiting to find out if you're blocked.
02
Every request gets a clear answer
Yes, not now, or no. Always with a reason. No ambiguity about where things stand or what comes next.
03
Exceptions ship with a note
Edge cases get a temporary exception and a note. The system stays intact. The team ships on time.
Outcomes
Custom tables 3 → 1.
A11y defects down 62%.
Data Table coverage 90%.
Sources
Component audit before and after rollout · A11y audit pre/post DS adoption · Screen audit, Hypotheek, rollout phase 1
Reflection

What earned trust. What I should have prevented.

What earned trust
What I should have prevented
The reframe was the project
"We can't conform" told us what the system was missing. Listening first turned a hostile team into a design partner.
Engaged Hypotheek six months too late
Should have been embedded with them from week one of the DS. By the time we showed up they'd already built three custom tables. Resistance is cheaper to prevent than to reverse.
Ship the contract before the components
The swimlane was the artifact that won them over. Architectural clarity beat visual polish.
Got Hypotheek, missed the room
While I was earning one team's trust, three other teams were quietly building around the system. I was so focused on the hardest team that I stopped paying attention to the easy ones drifting away.
Experiments

Things I build when nobody asked.

Tussenin
A Dutch practice workspace built with Claude Code, where every mistake becomes the next drill.
AI-nativeCorrection memoryLanguage learning
↗
Taalbits
Dutch grammar practice. Built to pass NT2. It worked.
Dutch learningBrowser onlyNo backend
↗
Designing an elderly homecare plan app — a UX case study
How We Empower Elderly Caregivers In Indonesia
Published by UX CollectiveHealthcare UX
↗
More on GitHub →
Let's work together
Richard
Andhika
chardrizard@gmail.com
Frequently asked
Do you need visa sponsorship?+
No. Dutch citizen, full EU work authorization. No sponsorship needed.
Are you open to working on-site?+
Yes. Based in Utrecht, well-connected to Amsterdam, Den Haag, Eindhoven, and Rotterdam. Hybrid is my natural default, happy to negotiate the on-site split. Face-to-face is preferred for collabs.
What's your availability?+
1-month notice period.
Do you work well in organizations without an existing design system?+
Starting from scratch without a clear brief is familiar territory, sometimes preferable. Chaos can be fun if you're enabled.
Are you a solo IC or do you prefer managing/leading a team?+
Strongly IC. I do my best work owning a problem end-to-end. I like coaching and giving direct feedback, but managing headcount isn't where I want to spend energy.

The craft side matters to me: I still want to be close to the work, doing the research, making the calls, pushing the visual quality. What also genuinely excites me is directing LLM agents as part of that workflow. I set the brief, the system executes, I review and steer. That's the kind of leverage I'm after.