Case 01 · Halliburton · 2025–2026

Data Marketplace

Shipping Halliburton's first unified data platform through an AI-native design-to-code pipeline.

Role
Lead UX, sole designer
Core team
3 engineers, PO, AI lead, Technology Director
Research
10 participants across 4 roles
Timeline
2025–2026
Status
In active use + testing

NDASensitive data and metadata are under NDA. All values are placeholders; process and design are shown in full.

Data Marketplace — screens across desktop and mobile
01

Problem & Context

Data assets live across disconnected systems with inconsistent metadata and unclear ownership.

Problem

Scattered and unfindable

Users cannot efficiently work with data because it is difficult to find, difficult to access, and not easy to compare across multiple sources.

Goal

One entry point

A unified marketplace for curated data, reports, models, and AI agents — part of Halliburton's Data Ribbon.

02

Roadmap

  1. 1.0

    Discovery

    Aligned with the Technology Director and AI lead on vision, then split the work into two phases — MVP and Trust/AI — so the design system could scale without rework.

    2 strategies · 1 foundation
  2. 2.0

    Research

    Mapped 4 persona groups, 5 user flows, user journey maps, and the information architecture that later powered the MCP hand-off.

    4 personas · 5 flows · IA
  3. 3.0

    Prototype & design system

    Started with AI-generated prototypes to validate structure fast, then built a scalable design system with reusable tokens and components.

    AI prototype · Figma system
  4. 4.0

    Usability testing

    Two rounds of testing with real users — Data POD Leads and Product Owners. Caught blind spots before a single line of code was written.

    2 rounds · 10 users
  5. 5.0

    Hand-off to dev

    Connected dev team to Figma through an MCP server. Developers pulled specs, tokens, and components directly into their IDE — no screenshots, no outdated docs.

    Figma MCP · live specs
  6. 6.0

    Post-launch research

    Tracked adoption, task success, and friction points after launch. The platform keeps evolving based on real usage data.

    Analytics · iteration
03

UX Strategy

Vision
Provide a user-friendly means of accessing curated, trusted data, reports, models, and AI Agents that are a part of Halliburton's Data Ribbon.
Success looks like
Data engineers use the dashboard daily as the primary entry point to find, validate, request access, and consume data assets within minutes — not days or weeks.
Persona groups
G1Data Scientist / Engineer
Browse data products, fast and relevant search, preview data, request & obtain access quickly.
G2Business User
Find existing reports across BI tools that answer questions from trusted data.
G3Data Product Owner
Manage data product info, approve/reject accesses, define level of access.
G4Data POD Lead
Oversee data products across teams, monitor quality and access patterns.
UX KPI

Efficiency — reduce

Search timeweeks → under 5 minutes
Access request timemonths → days

Adoption — increase

  • Engineers actively using the marketplace
  • Data products onboarded
  • Repeat usage rate

Quality & satisfaction — increase

  • User satisfaction score (CSAT)
  • Users who find the right product on first search
UX Goals
  • Help users find the right data product quickly
  • Make data product cards and details easy to understand
  • Simplify the access request flow
  • Make access statuses and permissions clear
  • Differentiate View and Edit access levels clearly
  • Enable product owners to add and manage products with minimal friction
UX Principles
  1. Reduce cognitive load — engineers already drown in data; the UI shouldn't add noise.
  2. Trust over decoration — every element earns its place; metadata, status, and lineage come first.
  3. Speed as a feature — every flow optimized for “minutes, not weeks.”
  4. Desktop-first — users live in multi-window data environments.
Core user scenarios
Search and discover relevant data productsFilter and narrow down data productsOpen dataset detail page and understand if it's usableRequest and get 1-click access directly from dataset pageTrack access request statusCompare data setsAdd and manage data productsManage access levels
Key deliverables
UX strategy for 2 product phasesPersonas (4 groups)User journey mapsJobs-to-be-doneUser flows (5 flows)Wireframes and mockupsInteractive prototypeScalable design system2 rounds of usability testing reportsPost-launch research report

NDABusiness goals and key features stay under NDA.

04

Research

Personas — 4 groups

G1

Data scientist / engineer

WhoTechnical specialist working with Python, ML models, and data integrations.

PainHard to find datasets across sources; long approval cycles; unclear ownership.

G2

Business user

WhoAnalysts, PMs, and business stakeholders using BI tools.

PainToo much raw data, too few curated business-ready reports; hard to trust freshness.

G3

Data product owner

WhoProduct manager responsible for governance and lifecycle of data products.

PainFragmented ownership; too many manual approval steps; slow onboarding.

G4

Data POD lead

WhoTeam/Tech Lead for a data engineering POD or domain.

PainHeavy manual documentation work; poor visibility into who uses their products.

User flows — 5 flows

Group 1 · Data scientist / engineer
  1. StartOpens the web app
  2. Step 1 · Sign in
  3. ScreenLogin page
  4. ActionEnter credentials, pick a rolethe role is asked once, on first sign-in
  5. SystemValidate and sign in
  6. Step 2 · Find the data
  7. ScreenAll data productsthe whole catalogue, unfiltered
  8. DecisionThree ways in
  9. Search
    1. ActionSearch by name or question
    Filter
    1. ActionFilter downdomain, owner, certification
    Browse
    1. ActionDrill into the catalogue
  10. DecisionAnything that fits?
  11. Step 3 · Judge and request
  12. ScreenData product pageschema, lineage, quality, owner
  13. DecisionCertified and trusted?the question the whole page has to answer
  14. ActionRequest access
  15. SystemCheck sensitivity and permissions
  16. Step 4 · Track the request
  17. ScreenTrack access requestsevery request and where it stands
  18. SystemEmail on the owner’s decision
  19. Approved
    1. EndAccess granted
    Rejected
    1. Off-flowReason shown, back to the catalogue

User journey maps — 4 groups

05

Prototype & Design System

Three-tier token architecture

One system, two themes, zero duplication. Every color flows through three layers — primitives, semantic tokens that swap for light/dark, and component tokens.

Hover or tap a token to follow it

~158 primitive · ~45 semantic · ~459 component tokens across 2 themes.

06

Usability testing

Before moving to development, I ran two structured feedback sessions with the primary user groups — validating designs, aligning on roles and workflows, and catching blind spots early.

Session 1

Data POD Leads

Design review with Data POD Leads — the technical owners who create and maintain data products. Validated roles, publishing workflows, and access management.

Key takeaways
  • Overall Marketplace concept is supported
  • Clear separation between POD Lead and Data Product Owner responsibilities
  • Automation-first approach validated
  • Access management should be user-centric
  • Sensitivity column removed from UI
  • Dev access management remains an open exploration topic
Session 2

Data Product Owners

Design review with Data Product Owners — the business owners responsible for final approval and publication. Validated publication flow, access governance, and role alignment.

Key takeaways
  • Overall Marketplace concept is supported
  • Clear separation between technical and business ownership
  • Access management should be user-centric with clear access reasons
  • Publication and access decisions should stay lightweight
  • Marketplace enables confident decisions without exposing technical complexity
07

Hand-off to Dev

Traditional hand-off means screenshots, redlines, and a PDF that's outdated the moment you save it. Instead, I connected developers directly to the Figma design system through MCP (Model Context Protocol) — they pulled specs, tokens, and interaction states straight into their IDE in real time.

How it flows
Figma fileDesign system, tokens, variables
MCP serverStructured data layer
Claude Code / IDELive queries from dev environment
Production codeComponents built 1:1 with design
08

Post-launch Research

What we'll measure
Adoption
Active users, onboarding completion, return visits.
Task success
Time-to-find, search-to-access, first-attempt access.
Satisfaction
Micro-surveys and interviews triggered in-product.
Pain points
Friction from session replays, tickets, heatmaps.
Modern research stack
Analytics copilot
Natural-language queries across usage data.
AI session replay
LLM insights instead of hundreds of recordings.
Continuous interviews
Weekly short calls, built in, not one-off studies.
Feedback synthesis
Surveys, tickets, and interviews themed by LLM.
Feedback-driven iteration
  1. ListenSurveys, replays, interviews, support tickets.
  2. SynthesizeTheme findings, prioritize by impact.
  3. ShipFix the worst friction in short cycles.
  4. ValidateRe-measure to confirm the metric moved.
09 — Outcome

Built, validated with real users, in active use while testing continues.

In place: a three-tier design system, an MCP pipeline wired into dev, and Figma-to-code hand-off.

10 — Reflection
I'd bring data governance in earlier — their requirements landed mid-project and sent the access flow back for rework.
Leading alone taught me seniority is owning decisions, not headcount. AI speeds up only what you already understand.