NorthStar program management platform on a MacBook

Amazon

NorthStar

A 0→1 enterprise platform that replaced fragmented workflows with one system for tracking work, sharing updates, and reporting project health.

0→1 Enterprise Design system
NorthStar program management platform on a laptop

Project details

Brand
Amazon
Role
UX Designer
Timeline
2024–2026
Platform
Desktop
Team
1 Designer (me), 2 PMs, 1 TPM, multiple engineers

For NDA confidentiality purposes
This case study uses generalized language, and mock data to protect proprietary information while preserving the design process and outcomes.

TL;DR

The problem

Every team ran program management its own way, in its own preferred tool, causing delays, bad data, and leadership decisions slowed by multiple report formats.

What I did

I ran 50+ interviews, mapped the jobs to be done, and rebuilt the information architecture from scratch, keeping the design aligned to the org-wide roadmap.

Timeline

Amazon, 2024–2026. I led the UX for the redesign, working with the product and engineering teams behind the tool. The MVP shipped in twelve months.

What changed

CSAT rose from 3.5 to 4.38. Monthly users grew steadily from 1,000 to 4,000 in three quarters. Workflows take fewer clicks, and Excel workarounds disappeared.

UX impact

0→1 MVP launched in 12 months
4.38/5 CSAT after 3 quarters (up from 3.5)
1K→4K Monthly users within three quarters of launch

The problem

Every team ran program management its own way, in its own tool, causing delays and bad data, so we needed one system to track work, share updates, and report project health.

What I did

My job was to turn that fragmentation into one system, without forcing every team into a workflow that didn't fit how they worked.

I led design from discovery to launch, shaping one system for projects, updates, and reporting across three user types.

  • Co-led research and workshops with product managers
  • Mapped user journeys and workflows
  • Prototyped and validated concepts with users
  • Shaped MVP scope as priorities shifted, alongside product and engineering

UX Methods

I started where I always start: with the people doing the work, in their actual context.

Stakeholder management

NA, EU, emerging countries

50+ interviews

Remote and in person

Group sessions

Co-led with product managers

Contextual inquiry

Users shared their screen

Competitor analysis

Pain points

Across 50+ interviews and three regions, the same four problems kept surfacing. We decided to build internally. That build became NorthStar.

Manual repetitive updates

Teams copied the same status across multiple tools with no single source of truth, creating unnecessary rework every cycle.

Terminology mismatches

Different teams used different language for the same work, making unified reporting to leadership inconsistent and unreliable.

No dependency visibility

When one team's work unblocked another, there was no way to know. People found out by asking around or missing deadlines.

Disconnected tools

Users toggled between several platforms to complete one workflow, creating friction and scattered, inconsistent data.

Synthesis & user outcomes

Pain points tell you what hurts. They don't tell you what people are trying to accomplish.

To better understand what users were ultimately trying to accomplish, I mapped their needs through a Jobs to Be Done (JTBD). This helped shift the focus away from features and toward the real outcomes users needed to achieve across planning, coordination, and execution.

NorthStar jobs to be done framework

Mapping the persona

Three very different users came out of that work, and the product had to serve all three: the admins who run the system, the project managers who live in it, and the leaders who only read the health of it.

Three personas shaped the product:

  • Central program managers Primary persona
  • Project managers
  • Leaders/viewers
NorthStar user persona

Understand the user journey

Then I followed the primary persona through one full cycle to find the exact moments where the old process broke down and where the new one had to hold.

After user research concluded, and we mapped the personas, we dived deep into the user journey to understand the gaps our new experience needed to fill.

The example below shows the primary persona's user journey.

NorthStar user journey map

We learned the outcomes. We understood the users. Now we needed to know how to build the experience.

Navigation and site map

With the journey mapped, the structure came next: one home for every project, and every critical workflow within one to three clicks of it. NorthStar had to serve three very different user types while still handling the complexity of enterprise program management.

The navigation itself never changed between roles: same bar, same sidebar, same order. What changed was how much of it a person could see. Each persona only saw the items their job needed, so nobody scanned past tools they could not use. Less surface, less cognitive load.

The house analogy

To communicate the IA structure across the team, I used a house analogy, where pages are rooms, actions are functions, and components are furniture. Every room exists for a purpose.

This framework helped:

Stay aligned On why each page existed during MVP definition
Prevent scope creep By questioning every new page against its purpose
Explain UX decisions To leadership and engineering in plain terms, which helped get buy-in early
NorthStar house analogy: rooms are pages, functions are actions, furniture is components

Navigation depth goal

I kept every critical workflow within 1 to 3 clicks of the home page, so users could find a project, timeline, or task list without digging through menus.

NorthStar navigation depth goal: users find what they need within 1 to 3 clicks

Permission-based navigation

The site map came straight out of what we learned. The interviews and the jobs to be done showed the three personas were not doing the same work at different depths, they were doing different work. So instead of one tree trimmed down per role, each persona got a structure shaped around their outcomes: admins configure, project managers execute, and leaders read. One home, one order, but nobody carried navigation their job did not need.

NorthStar navigation across personas

The opportunity was to create a unified system that improved visibility, reduced friction, and still supported the needs of different users.

Wireframes

This is what I’m allowed to showcase.

Note: All Amazon confidential data has been replaced with mock data for these designs.

NorthStar home screen
Home page

The landing page, a high-level view of every program and its status.

NorthStar project details sidesheet
Project details (sidesheet)

A quick way to view a program's details without losing context.

NorthStar project details page
Project details page

A deep dive into a program's deliverables, dependencies, and timelines.

NorthStar create new project with AI assistance
Create new project (AI assistance)

AI assistance helps fill out the new project form.

We built.
We measured.
We learned.

Results and impact

  • More consistent workflows

    Bringing scattered tools into one platform gave teams one structured way to manage projects and share updates.

  • Higher adoption

    Monthly users grew from 1,000 to 4,000 within three quarters of launch.

  • Clearer reporting

    Standard dashboards made it easier for admins and leaders to see progress, risks, and project health.

  • Improved satisfaction

    Post-launch feedback helped uncover usability gaps early. CSAT rose from 3.5 to 4.38 out of 5 over the first three quarters.

Lessons learned

NorthStar taught me that enterprise design is mostly about alignment. The hard part was balancing multiple user types, shifting product direction, and business priorities, then landing on something the team could realistically launch.

It sharpened how I define an MVP under pressure and use research to guide decisions. It also showed me that some of the most important improvements happen after launch, once real usage shows where the experience needs to change.