Variable 01
Problem type sets the route
Net-new features need ethnographic depth. Compliance updates need legal collaboration first. Performance issues need quant validation. Same designer, three different processes.
Adaptive Framework
How I Work:
How I Work:
Process & Methodology
Five years across UHNW real estate and regulated trading platforms taught me one thing: process depth follows risk. The depth of research, prototyping, and validation adjusts to what's at stake — not to a fixed playbook.
100K+
Traders served
40+
Jurisdictions
0
UX-tied findings · 4 yr
6
Defensible principles
Risk · Process Depth · Validation Rigor — calibrated together, not separately
The Adaptive Variables
Choosing the Right Route
Three dimensions decide how deep the process goes: what kind of problem, how much risk, how much time. Mismatched depth is the most common process failure.
Variable 02
Risk level sets the rigor
High-risk features get Legal in the wireframe room with parallel audit cycles. Medium-risk runs Maze. Low-risk visual changes ship and measure post-launch.
Variable 03
Timeline sets the breadth
Days mean reusing design system patterns. Weeks mean targeted custom design with one validation loop. Months mean full research → IA → prototype → validation.
Systematic Decision Logic
From intake to ship — four routes
Each route triggers different stakeholders, validation rigor, and handoff fidelity. Same designer, calibrated process.
Design Principles
Six Principles I Can Defend Under Questioning
Not mission statements — positions I've actually argued for in real meetings. Each one comes with a counterargument I had to talk past.
Principle 01
Whitespace signals different things to different audiences
Density signals competence to traders. The same whitespace signals discretion to UHNW clients. Applying one rule to both is a category error.
Principle 02
Legal-first is faster, not slower
Inviting Legal to wireframe review feels like friction — it's about 3 hours. Building 4 weeks of high-fi before compliance review creates the conditions for 4 weeks of waste.
Principle 03
Trust is earned through specificity, not polish
Expert evaluators read a wrong technical term as incompetence faster than they read a font choice. "Beautiful but vague" fails this audience. "Specific and accurate" survives.
Principle 04
Design for the most restrictive jurisdiction first
The compliance floor is the design ceiling. Starting permissive and retrofitting strict creates architectural debt that's nearly impossible to pay down without a full redesign.
Principle 05
Error prevention outranks efficiency for irreversible actions
A mis-submitted trade isn't recovered by undo — it's recovered by documentation, dispute, sometimes regulatory reporting. Confirmation friction is always worth it.
Principle 06
Cognitive context is the most expensive thing to rebuild
An interruption's cost isn't the 5-second pause — it's the 3-minute context reconstruction. Persistent panels beat tabs. Inline alerts beat modals.
Stakeholder Management
Design Decisions Are Political Decisions
Across Christie's and ACY (Sydney · Taipei · Jordan · Colombia), I learned a methodology must accommodate power dynamics — Legal veto, CFO scrutiny, UHNW expectations — not pretend they don't exist.
Legal Counsel
In the room from day 1
"Recommended Providers" was pixel-perfect when Legal killed it for ASIC inducement. Joint wireframe kickoffs since. Interactive prototypes, not static mockups.
CEO & CFO
Strategic purpose, not UX rationale
C-suite stakeholders need design serving a strategic purpose they've already defined — investor materials, capital-markets reporting, fiduciary-weight data viz.
UHNW Clients
Editorial-first, not self-serve
$50M buyers expect personal broker treatment. Properties presented like Architectural Digest features. Digital tools enhance broker relationships — they don't replace them.
Cross-Cultural Teams
Async-first, language-aware
Sydney HQ + Taipei dev + Jordan/Colombia satellites = 4 timezones. Bilingual specs (English UI + Chinese rationale) eliminated the lost-in-translation tax.
Compounding Knowledge
How Each Project Changed the Next
Not a list of isolated work. Each one changed the questions I asked next: ACY (production, regulated) → TradeX (conceptual, deeper domain) → Xanthos (different user class, same regulation class).
Arc 01
ACY taught me that data density is a user-class requirement
Watching experienced traders close my "simplified" UI and revert to MT4 clarified TradeX's thesis: institutional users don't need UX to manage cognitive load — they've trained for it.
Arc 02
TradeX clarified the persona model that made Xanthos sharper
Three users · same data · different views. The principle is the same: when two roles share a product, their relationship should be visible in the UI, not hidden behind role-switching.
Arc 03
Christie's permanently changed how I design for wealth
Three private auction previews. Fewer than 10 conversations. They are unsentimental about information, do not want to be educated by the interface, and evaluate institutions like art.
Research Framework
How I Structure Research for Financial UX
Constrained by NDA, regulatory sensitivity, and users who evaluate your research competence alongside your design. This four-phase framework is tested across 5+ studies in 4 years.
Phase 01
Map constraints first
Before writing the interview guide: what's locked (regulation, architecture)? What's actually open? Researching locked space wastes expert practitioners' time.
Phase 02
Domain-primed interview protocol
PMs, quants, compliance officers are experts. Generic "show me" produces surface findings. Open with the specific problem space; ask about failure modes, not task completion.
Phase 03
Demand categories, not personas
"Sarah uses the risk module" isn't actionable. "Needs cross-asset correlation during a news event" is. Demand categories stay stable across compliance-driven redesigns; personas drift.
Phase 04
Calibrated confidence claims
Directional findings are useful for ruling out failure modes — not for validating interaction patterns. Reporting wrong confidence is a methodological error that erodes credibility.
What Doesn't Work
Three Failures That Taught More Than Ten Successes
Process maturity isn't always being right — it's recognizing failure signals early and pivoting before sunk costs become catastrophic.
Failure 01
Over-engineering research for low-risk changes
PM was furious: "We could have A/B tested this in 4 days with 100K+ users." Treating every problem like it needs deep research is "user-centered" theater.
Failure 02
Assuming stakeholder alignment without validation
Designed "Recommended Providers" on PM requirements alone. Legal killed it for ASIC inducement. Now Legal sees clickable prototypes before pixel-perfect handoff.
Failure 03
One-size-fits-all handoff (cross-cultural fail)
Taipei team read Figma specs literally but missed the design rationale Sydney heard verbally. Bilingual specs (EN labels + ZH rationale) + Loom walkthroughs lifted match 60% → 95%.
Active Learning
What I'm Studying to Close the Gaps
Naming knowledge gaps precisely is itself a design skill. Three areas I'm actively closing as I move from retail toward institutional finance.
Learning 01
Bloomberg Market Concepts
The gold standard for institutional terminals. Understanding function-key IA, multi-monitor workspaces, and FXFM/PORT/ALLQ workflows informs my institutional design vocabulary.
Learning 02
UHNW behavioral finance research
Supplementing direct observation (3 Christie's previews) with Capgemini WWR and UBS Investor Watch — quantitative patterns behind what I observed qualitatively.
Learning 03
Trust structure & multi-entity UX
Private banking operates at 3–4 entity levels (client → family office → trusts → portfolios). Studying entity-switching IA, role-based access, cross-entity KYC/AML triggers.
Why I share gaps openly: I joined retail brokerage with no FinTech background and was leading compliance-architecture decisions across 40+ jurisdictions within 18 months. Not because I learn fast — because I translate domain structure from first principles. The gaps above are specific and bridgeable.
Tools & Handoff
Facilitating Shared Understanding
The toolkit isn't about drawing — it's about making intent legible across distributed teams with competing priorities and regulatory constraints.
Design
Figma, ProtoPie, Framer
Research
Hotjar, Maze, UserTesting
Handoff
Storybook, Zeroheight, Loom
Data
Mixpanel, Google Cloud Console
Performance handoff: Canvas vs SVG decision
In institutional finance, "jank" isn't a minor annoyance — it's a data integrity hazard. I define rendering strategy alongside engineering, not after.
SVG (standard B2B)
Static reports, slow analytics. Accessible DOM, easy styling. Limit: degrades at >1,000 nodes / 60fps.
Canvas
(institutional)
High-frequency grids in LogixTrader / TradeX. Fixed memory footprint, constant-time render. Trade-off: custom screen-reader implementation.
I provide the JSON schema for Canvas-based "High-Performance Data Tables" that hold 60fps during volatile news events. Engine constraints, not just mockups.