Case two · J.P. Morgan · 2023 to 2025 · shipped

Fusion analytics dashboard

Portfolio managers and analysts were drowning in fragmented data. Their real need was trust in the numbers, not more features. Performance, attribution and risk they could take into a board meeting.

Role · Digital Product Designer VP. Led the designSystem · Salt, plus a Fusion layer I builtStatus · Shipped, opened to clientsLive · My public rebuild (opens in new tab) →
Original Fusion Analytics Dashboard design with report discovery, featured dashboards and the report table
Original Fusion Performance design showing the complete workspace

01 Problem

  • Users

    Portfolio managers and analysts at asset managers, and in phase two their clients, who'd had no training. Expert users under time pressure, often presenting the screen live.

  • The job

    Performance, attribution and risk they could take into a board meeting, from live and historical data, with heavy filtering.

  • Where it started

    The existing RaaS interface was a legacy layout that couldn't flex to different data or roles.

User personas

My three design personas: the portfolio manager who presents, the analyst who slices, and the client who never had training. Composite roles from the case, not research participants.

Helen Marsh

Portfolio Manager Asset manager, presents live

I need numbers I can take into a board meeting.

Context

Prepares performance, attribution and risk for board meetings, and presents the analysis live.

In this app
  • Board-ready analytics on Salt
  • Type, hierarchy and colour set for the worst-case viewing moment
  • Configurable pages built from widgets
Goals
  • Performance, attribution and risk for a board meeting
  • Trust in the numbers, not more features
  • Present the screen live, under pressure
Pain points
  • Drowning in fragmented data
  • A legacy RaaS layout that couldn't flex to different data or roles
  • Dense data pulling against decision speed

Tom Reilly

Investment Analyst Returns, attribution and risk

Select a cube, slice it, and get to the view that matters fast.

Context

Explores returns, attribution and risk across data cubes to build an analytical view.

In this app
  • Advanced filtering, toggles and deep slicing
  • One widget serves many cubes
  • Import and export through FactSet
Goals
  • Select a data cube and slice it deep
  • Analyse returns, attribution and risk across asset classes
  • Report without leaving the surface
Pain points
  • Heavy filtering across live and historical data
  • Three levels of configuration, the middle one added load
  • Every new data cube would have needed a new page

Amara Osei

Client, phase two Had no training

Nobody trained me. It has to make sense on first open.

Context

Uses the client-facing view in phase two, without the training available to internal analysts.

In this app
  • Entitlement-based views
  • Same screens for internal and external users
  • One codebase, consistent patterns
Goals
  • See the part of the book she is entitled to
  • The same patterns as the internal team
  • Get to a useful view without onboarding
Pain points
  • No training before first use
  • A simplified client version would drift from the product
  • Permissions differ by person

02 Constraints and role

  • My role

    A UX designer was shared across other projects, so most of the day-to-day calls were mine. A product owner set requirements.

  • Salt, the firm's design system

    Every component had to come from Salt or go back into it. I built the Fusion layer on top and contributed back. My craft shows in layout, hierarchy and data density, not in new components.

  • Adoption was part of the problem

    Some development teams built custom components instead of using the design system's components out of the box. The UI became inconsistent and harder to extend and evolve. I audited the custom components, organised them into groups and reviewed the issues regularly with developers as we worked through the fixes.

  • FactSet's engine behind the data

    The analytics came from a partner engine, so the shape of the data was fixed. Design had to fit what the engine could return, fast.

  • Accessibility as a gate, not a finish

    An accessibility specialist sat in every review. Anything that failed went back before it reached engineering.

  • Two constraints that pulled against each other

    Dense data and decision speed. I named them both, then built the conflict into the design. Type, hierarchy and colour were set for the worst-case viewing moment, someone presenting live under pressure, not the clean demo.

03 Research and workflow

User flow

An analyst shapes the workspace, checks the view, then presents and reuses it.

Case-study workflow · entitlement-based access and analytics in the same surface.

On smaller screens, enlarge the flow to read and explore the branches.

Fusion · from data to a board-ready view. An analyst shapes the workspace, checks the view, then presents and reuses it.Fusion · from data to a board-ready view. An analyst shapes the workspace, checks the view, then presents and reuses it.
  • Sketch, wireframe, high fidelity

    Each flow went through paper, then wireframes, then built screens on Salt. The sketches set the information hierarchy before any visual decision.

  • The review cycle

    Requirements with the product owner, iteration with UX, review with product, engineering and the accessibility specialist, then pixel-level work with developers in daily stand-ups. Accessibility was a review step, not a final check.

  • What testing changed

    The customisation tier I cut came from watching people use the configuration options. The middle tier added choices without changing what they built. Removing it made the first useful view faster to reach.

User journey

Helen Marsh · Portfolio Manager · Asset manager, London · Quarterly board pack, one afternoon

Persona

Helen Marsh

Portfolio Manager Asset manager, presents live

Scenario

Helen is preparing a quarterly board meeting. She opens Fusion to pull performance, attribution and risk for the book, slices by asset class and period, and builds the view she will present live while the numbers are still updating.

Goal

Walk into the board meeting with numbers she trusts, built in one surface, without a spreadsheet detour.

Start with the book

01Find the data

Helen selects the data cube for her book. FactSet's engine feeds it, so the shape of the data is fixed and the first useful view has to be fast.

Design response
  • Entitlement-based view on open
  • Cube picker with recent cubes
What the system does

FactSet engine returns the cube. The shape of the data is fixed, so the first view has to be fast.

Refine the view

Filters and analysis stay in the same workspace.

02Shape the view

Filter and slice: asset class, period, benchmark. The old RaaS layout couldn't flex to different data or roles.

Design response
  • Advanced filtering, toggles, deep slicing
  • Comparison tools tailored to data complexity and role
  • Two configuration levels, not three
What the system does

Salt filter bar and toggles. Entitlements decide which options each person sees.

03Analyse

Returns, attribution and risk across asset classes, in widgets that work across cubes and permission levels.

Design response
  • Modular widgets
  • Data tables and chart frames on Salt
What the system does

One widget shell renders a table or a chart from the same data contract, across cubes.

Back to filters when the view needs to change.

Take the view into the room

The audience changes. The data stays in the same surface.

04Present

She presents the screen live, under pressure. Type, hierarchy and colour were set for this moment, not the clean demo.

Design response
  • Board-ready report without leaving the surface
  • Worst-case viewing moment as the baseline
What the system does

Dashboard page configuration saved per user. Accessibility specialist reviews before release.

05Share and reuse

Phase two: the same screens reach external clients who had no training. One codebase, entitlement decides what each person sees.

Design response
  • Import and export through FactSet
  • Same patterns for internal and external users
  • Patterns reused across other Fusion modules
What the system does

One codebase. Entitlement-based views for clients. Import and export through FactSet.

Opportunities

Get in front of end users directly and earlier. The start leaned on product-owner requirements. Ship a leaner version one with fewer modules, and keep the accessibility specialist involved from the first review.

04 Decisions and trade-offs

Decision map

The dashboard had to support different data, different decisions and different permissions. These calls address reuse, configuration effort and the move to external clients.

On smaller screens, enlarge the map to read and explore the branches.

Fusion decision map. One analytics surface, from internal teams to external clients.Fusion decision map. One analytics surface, from internal teams to external clients.

Reuse without repeated setup

Widgets let the same structure serve different cubes. Removing the middle configuration tier limits how much setup that flexibility asks of the user.

One structure, different access

The same page and widget patterns then support external clients through entitlements, with review of what each audience can see.

  1. Shipped and reused

    01Dashboard structure

    Should each report have its own page, or share a widget structure?

    Performance, attribution and risk draw on different data cubes. Fixed pages make the first report straightforward but make each new combination another layout to maintain.

    I chose

    Modular widgets
    Modular widgets

    Reuse widgets across data cubes and permission levels.

    Alternative

    Fixed layouts per report type
    Fixed layouts per report type

    Build a fixed page for each report type and data combination.

    Why I chose it

    We explored fixed layouts per report type and a widget grid. Fixed layouts were faster to ship, but every new data cube would have needed a new page. The grid won because one widget serves many cubes and permission levels.

    What it cost

    Slower to ship than fixed layouts, and every widget had to earn its place in one of the four workflow steps.

    What informed the decision

    I watched internal users configure pages during weekly reviews on their own portfolios and used their choices to shape the default layout. Other Fusion teams later reused the widget and filtering patterns.

  2. Simplified after review

    02Configuration

    How much configuration actually changes the view someone needs?

    The first design offered page, section and widget settings. The middle level added another place to look before reaching the same result.

    I chose

    Cut a customisation tier
    Cut a customisation tier

    Keep page and widget configuration; remove the section tier.

    Alternative

    Three levels of configuration
    Three levels of configuration

    Retain three levels and the extra choices for each level.

    Why I chose it

    Early versions had three levels of configuration. Testing showed the middle tier added cognitive load without changing what people built. I removed it. The test I use for what not to design: is this serving a real workflow contradiction, or one team's preference?

    What it cost

    Removing a level also removes a place to configure a section independently. I kept the page and widget controls that supported the useful views.

    What informed the decision

    I watched people use the three levels during the same reviews. The middle tier added choices without changing what they built. This was not a timed study, so there is no measured speed improvement.

  3. Opened to clients in phase two

    03Phase two

    Should external clients get a separate product?

    Phase two brought in clients without training. A second interface would add another set of patterns to maintain; a shared one had to respect their narrower permissions.

    I chose

    One product with entitlement-based views
    One product with entitlement-based views

    Keep a shared product and make the permitted view work for each role.

    Alternative

    A simplified client version
    A simplified client version

    Create a separate simplified client interface alongside the internal product.

    Why I chose it

    Phase two brought in clients who'd had no training. We chose one product with entitlement-based views over a simplified client version, so patterns stayed consistent and the team kept one codebase.

    What it cost

    Client constraints live in the same codebase as the internal views.

    What informed the decision

    The client-facing team informed the work, and two clients tested before phase two widened. That is a limited review record, not evidence that every client needed no support.

The current Fusion performance table beside its configuration panel

05 Design and system

The screens below are my original Fusion designs. The recording shows my public rebuild using sample data. FactSet integration and entitlement-based access describe the original project; those services are not connected to the rebuild.

Silent recording. It moves from report discovery into the Performance workspace, opens its configuration panel, then shows Sustainable Investment.
Original Fusion design with widget configuration and a custom date range
Original Fusion Performance design with performance results and analytical widgets
Original Fusion Risk design with portfolio risk charts, tables and analytical widgets
  • Before Fusion, I helped define the foundation

    I worked on UITK, now Salt, on the design language, the look and the feel. I ran the design system workshops in London and New York with design, product and development in the same room. The point wasn't to present. It was to agree what a component is before anyone built it. On Execute, I worked with developers from day one, weekly working sessions and daily when needed.

One foundation, a reusable Fusion layer

  • Salt → Fusion → widgets

    I built tokens, components, patterns and accessibility layers on Salt: data tables, chart frames, filter bars and widget shells. Gaps went back upstream. Other Fusion teams reused the same components and patterns.

  • Many cubes, familiar views

    FactSet supplies performance, attribution and risk data. The same widgets serve different cubes and permission levels. People configure their own dashboard pages; entitlements control what they see. Import and export run through FactSet.

I kept Fusion on Salt so it could follow the firm's updates, and contributed missing patterns back. The diagram describes the original product; the public rebuild is not connected to FactSet or entitlement services.

06 Outcome

  • Reviewed

    I worked closely with an ADA specialist on a specific flow and used developer tools to run accessibility checks. This describes the flow reviewed, not a product-wide audit or WCAG certification.

  • Adopted

    Widget and filtering patterns reused across other Fusion modules.

  • Phase 2

    Opened the product to external clients.

Original Fusion screening workspace showing screening changes and their status
Original Fusion Sustainable Investment design comparing two entities across ESG, emissions and controversies

Accessibility and responsiveness

Accessibility was a review step.

An accessibility specialist sat in every review. Anything that failed went back before it reached engineering.

Across screen sizes

The rebuild's catalogue reflows at 820px and at 390px. These captures show the catalogue. The dense analytics workspaces are desktop first, and I say so.

Current product at an 820-pixel tablet viewport
Tablet · 820px viewport
Current product at a 390-pixel phone viewport
Phone · 390px viewport