Hi, I’m Jade.
A visual storyteller and designer shaping calm, considered work — identity, editorial, motion, and the small things in between.
Selected work — Jade
Selected work
{{ project.blurb }}
{{ project.blurb }}
{{ cur.cat }}
{{ cur.title }}
{{ cur.summary }}
{{ cur.contribution }}
Product Design Case Study
RealSelf Provider Profile
How a profile became a higher-confidence decision path.
Patients evaluating elective treatments were navigating a dense mix of credentials, reviews, specialties, and before-and-after galleries — without a clear way to determine what mattered or when to book. I led the end-to-end redesign of RealSelf's provider profile, using behavioral data and user research to define a clearer information hierarchy and consultation path.
This was more than a visual refresh. I aligned Provider, Sales, Ads, and engineering around a shared decision model, balanced patient trust with business and technical constraints, and translated it into a shipped experience that increased Book Consultation CTA click-through from 45% to 65%.
Product Design Case Study
Event Travel Operations
How a feature became a platform.
{{ cur.summary }}
And I didn't just build it — I ran it like a shipped product: versioned, critiqued, and changelogged. Fifty-four documented decisions, each with a principle I can reuse.
Conservative estimates from observed pilot baselines.
Where patient trust becomes consultation intent.
RealSelf brings together the signals patients use to evaluate cosmetic-treatment providers: credentials, reviews, ratings, answers, and Before & After photos. The provider profile sits at a high-intent moment in that journey, where patients assess a doctor's credibility, relevance, and results before deciding whether to book a consultation.
For providers, the same surface must communicate expertise clearly, establish differentiation, and create a direct path from consideration to consultation.
Design intentRe-architect a fragmented profile into a coherent decision system — one that clarifies trust signals, strengthens provider visibility, and creates a more direct path to consultation.
One profile had to serve patients, providers, and the business.
Patients needed to compare credentials, specialties, recent reviews, expert answers, and clinical results — without clinical expertise or a standardized way to judge quality.
Providers needed a credible way to communicate their expertise, distinguish their practice, and make their strongest proof easier to discover.
At this high-intent point in the journey, information hierarchy, trust signals, and CTA placement became key design levers for Book Consultation engagement.
The profile contained the right signals—but not a clear decision hierarchy.
- 01Critical trust signals were distributed across a long, difficult-to-scan page.
- 02Provider expertise—including specialties, expert answers, and clinical results—lacked visual priority.
- 03Long scrolling paths made reviews, Q&A, specialties, and Before & After content difficult to find and revisit.
- 04The existing taxonomy did not consistently reflect the language patients used to search for treatments.
- 05Book Consultation CTA click-through was 45% at a high-intent point in the journey.
- 01Define a reusable information architecture organized around how patients assess credibility, fit, and results.
- 02Elevate provider expertise and make differentiating evidence easier to evaluate.
- 03Make high-value content easier to discover, navigate, and revisit.
- 04Align labels and treatment categories more closely with patient search behavior.
- 05Create a clearer path to the Book Consultation CTA and improve click-through engagement.
Original desktop
Original mobile
A content-rich profile without a clear decision hierarchy.
Attention became less concentrated deeper in the page.
Interaction was distributed across disconnected sections.
The content was present, but patients had to assemble credibility, evidence, and next steps across an extended page.
Research turned a broad problem space into a sequenced product strategy.
Frame the decision space
I interviewed partners across Provider, Sales, Ads, Product, and Engineering to map user pain points, prior learnings, business goals, and technical constraints. This defined the requirements and tradeoffs the redesign needed to address.
Turn behavior into hypotheses
In weekly working sessions with BI, I connected behavioral patterns to testable design hypotheses and defined the MVP learning agenda. Behavioral analysis and prior research pointed to three signals worth prioritizing and testing: review recency, provider specialties, and visual proof through Before & After content.
Validate and align
I ran a usability study with 10 participants, then facilitated a workshop with PMs, Sales leaders, and VPs to synthesize findings, pressure-test assumptions, and prioritize the strongest opportunities.
Sequence the strategy
The research led to a two-track strategy: improve how patients browse and compare provider evidence in the near term, while rebuilding the underlying information architecture for longer-term scale. Because the work intersected with technical dependencies and an upcoming React migration, I sequenced it as incremental releases rather than a one-time redesign.
Mapping page content to patient decisions
The original profile organized information by page sections. The new architecture reorganized it around how patients evaluate credibility, expertise, evidence, and next steps.
Original page structure
High-value signals were separated across an extended vertical stack.
Patient decision tasks
Decision-based architecture
The persistent summary stays visible while evidence is organized into scannable destinations.
Six decisions that turned a page into a decision system.
A persistent profile summary and task-based tabs reorganized a long content stack into clear destinations—letting patients move directly between expertise, reviews, results, and answers without scanning the full page.
Prioritized rating, review recency, post date, and quick-review actions so patients could identify relevant feedback more efficiently.
Highlighted treatment keywords, surfaced media counts, and prioritized questioner context to make answers easier to browse and evaluate.
Added verified status to the sticky provider bio, clarified back-navigation, and made the Book Consultation action more visible without disconnecting it from provider context.
Worked with a UX writer to move away from medical jargon and organize services around the terminology and needs patients use when searching.
Replaced the limited horizontal carousel with categorized, provider-generated photos in a waterfall-style gallery, making clinical results easier to explore and compare.
A clearer identity and consultation action moved the number that mattered.
Book Consultation CTA click-through rate. A clearer sticky provider identity and a more visible consultation action increased CTA click-through from 45% to 65%.
Time on page decreased slightly. On its own this is not proof of more or less engagement — read alongside the higher CTA click-through, it is a possible indication of more efficient navigation.
The result supported the redesigned profile structure as a stronger decision and conversion pattern, with potential application across other core RealSelf pages.
Designing clearer states and more accessible product interactions.
The redesign made key states more legible: a distinct verified identity, an explicit selected-tab state, and a high-contrast consultation action separated from surrounding content. Reorganizing a long stack into labelled tabs also gives each section a clear, reachable destination rather than a position in an endless scroll.
Described from the visible design evidence in the profile. No audit scores, certifications, or numerical accessibility claims are implied.
From a redesign to a shared product language
I used the Provider Profile redesign as a catalyst to align Design, Product, and Engineering around a shared system—translating one feature overhaul into reusable foundations for more consistent decisions, faster delivery, and a scalable product language.
System foundations
Applied across the product
Design intentOne library, two modes — browse as cards, compare as a full-width table.
Design intentCoordination never ends at setup — the interaction model had to survive what happens after creation.
Design intentSpeed for coordinators and guardrails for the org — without trading one for the other.
Design intentNew events start from the baseline; only the differences get configured.
Design intentThe roster resolves in parallel — the system surfaces only what needs human judgment.
Design intentLaunch is a risk review, not a setup step — going live is a deliberate, accountable decision.
Design intentOne live view of the cohort replaces status-chasing across tools.
Design intentCompliance by structure, not by chasing — every inheritance and override is captured.
One operating model, end to end.
From one feature to an operating model.
Events shipped. Then testing showed the same setup rebuilt across ~10 programs — so I evolved it into Specialty Desk: locked baselines, per-event configuration, auditable launches. It became a core product offering, running real enterprise travel programs today.
~80% less setup per event · one place instead of 3+ tools · every override audited · ~15 hrs/week returned per coordinator — conservative estimates from observed pilot baselines.
Owned the calls
Sole designer and de-facto PM: I framed the problem, chose the lifecycle model over hub and wizard, and aligned engineering and operations on the decision that mattered most — what stays locked.
Authored the abstraction
A template is defined by what you pre-fill. Every field carries its own authority — locked, override-with-approval, or free. Speed for coordinators, guardrails for the org.
Ran design like a product
Fifty-four documented decisions, versioned and changelogged, each distilled into a principle that constrained the next round.
- Travelers who bypass the guided flow from open search meet policy only at checkout — v2 detects event trips from search context.
- Per-traveler date variation inside one trip is still an open edge.
Founder · side project
JawiFit
JawiFit is my own fitness-studio side project. The app and web products aren’t launched yet — so this page shows the two things I designed end to end: the brand’s design system, and the Progress Wall, a consent-aware system for turning verified training progress into public celebration. Central principle: celebrate progress — not position.
Training systems are full of motivating progress — and full of reasons it can’t just be posted.
Coaches want to celebrate client wins. But raw training data is private, records are unverified, and public comparison turns motivation into rankings. The design question was never “how do we post progress” — it was “what makes progress safe to celebrate.”
Celebrate progress, not position.
The wall shows verified achievements, never rankings — comparison is designed out, not moderated out. Only opted-in, coach-verified, published achievements appear, with public-safe fields only.
One lifecycle, two independent eligibility axes.
My core structural call: evidence and consent are separate axes, not lifecycle states. The record moves needs-review → verified-private → published (rejected from any pre-publish state), while evidence (complete / incomplete / unresolved) and consent (eligible / missing / revoked) gate it independently. Consent is client-owned — the consent form is the source of truth; coaches get view-only status and a request-update action, never an override.
Verifying and publishing are two separate decisions.
Automated candidate detection with a manual fallback feeds an evidence-backed review queue — the coach’s primary surface. Every candidate carries its source, comparability checks, rule version, and confidence. A coach can verify a record without the power to publish it; nothing in the queue is public.
The visibility gate is explicit — it shows exactly what becomes public.
At publish time the private record sits beside the exact public card, with a consent check and a keep-private escape hatch. Publish only lights up when consent is eligible and the wording is safe — and a wording mistake is recoverable without ever losing the verified record.
Five red-team passes — I attacked it before anyone else could.
V4 shipped one clickable end-to-end slice; V5 to V5.4 then ran five independent browser-based red-team passes. They produced fail-closed privacy detection (any phone-like digit sequence, Unicode-aware), real dialog semantics with focus traps and inert backgrounds, 44px hit areas, reduced-motion support, and three verified viewports including a 1920 wall display.
No launched app to point at yet — so the discipline is the product.
Hi-fi prototype with fictional data; the full V1–V5.4 design history lives in the design archive.
About — Jade Yang
About
Hey, I’m Jade.
I’m a Bay Area product designer turning complex human workflows into calm, scalable product systems.
My path runs through law, dance, enterprise travel, marketplace trust, and founding JawiFit. Law trained me to reason through ambiguity; movement shaped how I think about behavior and experience; building a real fitness business keeps my work close to customers, growth, and operational reality.
I design B2B and B2C products with a bias for clarity, trust, and systems that actually work in the real world.
Designer at RealSelf
Graduate at Parsons
Undergraduate at China University of Political Science and Law
© 2026 Jade Yang