rado

Work

Available for work

Let's chat

About

Available for work

Let's chat

Work

About

rado

CASE STUDY | Password protected


SaaS data analytics -

rebuilding trust through clarity


SaaS analytics - rebuilding trust through clarity


SaaS data analytics -

rebuilding trust through clarity

Apex is Optum's healthcare analytics platform, used by GPs, Integrated Care Boards (ICBs), and Primary Care Networks (PCNs) across the UK to manage population health data, clinical performance, and commissioning insights. For most users, it's part of the daily rhythm of running a practice.


The platform had been live and evolving for years, and the homepage showed it. Features had multiplied, banners layered in with every release, navigation drifting as modules were added and reshuffled. No single change was the problem; the cumulative effect was. By the time we were brought in, users were telling us through support tickets, drop-off rates, and direct feedback that the dashboard wasn't working for them anymore.

Senior Product Designer. I led the dashboard redesign for Apex, one of several projects during my time at Optum. Collaborated with a cross-functional team of 1 researcher, 1 copywriter, 3 developers, 1 product owner, 1 project manager, 1 business analyst, and 4 support and sales members


Industry: Healthtech | B2B, SaaS, Data Analytics

Task delivery time: 2 months

IMPACT AFTER PHASE ONE

IMPACT AFTER PHASE ONE


70%

Improved feature discovery

70%

Improved feature

discovery

18 points

User satisfaction scores improved

38%

Orienting faster and acting

with more confidence

63%

Backlog reduction

63%

Backlog reduction

63%

Backlog reduction

47%

Increased employee productivity

47%

Increased employee productivity

47%

Increased employee

productivity

discovery phase

discovery phase

The breaking point


Apex's homepage had lost its sense of priority. Promotional content sat where operational content should have been. Critical alerts competed with marketing tiles for the same visual weight. Colour was decorative rather than meaningful, space was filled where it didn't need to be and empty where it did, and the information users actually came for was rarely the first thing they saw.


The result was a screen that asked users to do the platform's job for it. Every login began with a scan, a filter, a mental sort through noise to find the one or two things that mattered that day. For a tool people opened before morning clinic and returned to between patients, that friction added up fast.


Legacy APEX dashboard

The breaking point

The homepage had become a wall. Critical alerts were buried beneath promotional banners. Navigation patterns had drifted. Training resources were invisible. Users had no clear sense of where to start, and they'd stopped trying

Our support team forwarded a message from a practice manager:

—————-

"I log in, I look at the screen, and I honestly don't know what

I'm supposed to do next."

—————-

Legacy APEX dashboard

The breaking point

Our support team forwarded a message from a practice manager:

"I log in, I look at the screen, and I honestly don't know what I'm supposed to do next."

She wasn't alone. Users were opening APEX, feeling overwhelmed, and reaching for the phone instead of using the platform. The dashboard had become an obstacle.

The breaking point

The homepage had become a wall. Critical alerts were buried beneath promotional banners. Navigation patterns had drifted. Training resources were invisible. Users had no clear sense of where to start , and they'd stopped trying


Our support team forwarded a message from a practice manager:


"I log in, I look at the screen, and I honestly don't know what I'm supposed to do next."


Legacy APEX dashboard


A practice manager summed it up in a message forwarded by our support team:


"I log in, I look at the screen, and I honestly don't know what

I'm supposed to do next."



The breaking point

The homepage had become a wall. Critical alerts were buried beneath promotional banners. Navigation patterns had drifted. Training resources were invisible. Users had no clear sense of where to start, and they'd stopped trying

Our support team forwarded a message from a practice manager:

—————-

"I log in, I look at the screen, and I honestly don't know what

I'm supposed to do next."

—————-

Legacy APEX dashboard

The breaking point

Our support team forwarded a message from a practice manager:

"I log in, I look at the screen, and I honestly don't know what I'm supposed to do next."

She wasn't alone. Users were opening APEX, feeling overwhelmed, and reaching for the phone instead of using the platform. The dashboard had become an obstacle.

The breaking point

The homepage had become a wall. Critical alerts were buried beneath promotional banners. Navigation patterns had drifted. Training resources were invisible. Users had no clear sense of where to start , and they'd stopped trying


Our support team forwarded a message from a practice manager:


"I log in, I look at the screen, and I honestly don't know what I'm supposed to do next."


Legacy APEX dashboard

Constraints:

 2-month deadline for phase one. No design system. No time for large-scale research. Legacy architecture constrained what we could build. Communication across the team was fragmented.

Defining business goals


• Increase feature adoption from 20% toward 50%

• Reduce support tickets by 20%

• Improve satisfaction to strengthen retention

• Create a foundation for upselling advanced features

Defining user goals


• Restore trust through clarity

• Reduce time-to-orientation

• Increase engagement with key sections

• Reduce support dependency through inline guidance



How did we discover it?


Research insights


I analysed 50+ app store reviews and categorised complaints by theme. 60% of negative reviews came from confused new users. 40% came from existing users frustrated by poor navigation. The user flows told the same story: to buy Bitcoin, users had to tap the menu, scroll to "Trade," select "Buy," choose Bitcoin from a dropdown, then confirm.


8 interviews uncovered the pattern. AI helped me articulate it.


The researcher and I interviewed 8 users across three roles, GP practice admins, ICB managers, and PCN managers, testing six journey scenarios identified from BA documentation and sales team workshops.


The BA document mapped out the most common journey in APEX and the user research board.


Pulled the BA's data analysis document, compared what users actually did vs. what the platform expected


The researcher and I interviewed 8 users across three roles: GP practice admins, ICB managers, and PCN managers. We tested six journey scenarios identified from BA documentation and sales team workshops.


The BA document mapped out the most common journey in APEX and the user research board.

Research insights


I analysed 50+ app store reviews and categorised complaints by theme. 60% of negative reviews came from confused new users. 40% came from existing users frustrated by poor navigation. The user flows told the same story: to buy Bitcoin, users had to tap the menu, scroll to "Trade," select "Buy," choose Bitcoin from a dropdown, then confirm.


8 interviews uncovered the pattern. AI helped me articulate it.


The researcher and I interviewed 8 users across three roles, GP practice admins, ICB managers, and PCN managers, testing six journey scenarios identified from BA documentation and sales team workshops.


The BA document mapped out the most common journey in APEX and the user research board.


8 interviews uncovered the pattern.
AI helped me articulate it.

Six journey scenario modules were identified for surfacing:


• Access and demand

• Enhance access

• Critical notifications

• Clinician activity

• Reattendance

• ARRS roles


8 interviews uncovered the pattern.
AI helped me articulate it.

Six journey scenario modules were identified for surfacing:


• Access and demand

• Enhance access

• Critical notifications

• Clinician activity

• Reattendance

• ARRS roles


I used Copilot AI to transcribe and analyse interview recordings and parse BA documents, surfacing cross-role patterns faster than manual synthesis.

The transcripts revealed the same pattern. Users knew what they needed to do. They couldn't figure out where the platform wanted them to do it.


01

"I can't tell what's urgent." 

02

"Confusing dashboard, I don't know where to start." 

03

"I didn't know training resources existed." 

04

"I always have to search for help." 

The core insight: Users weren't struggling because APEX lacked features. They were struggling because the interface gave them no sense of priority. Everything looked equally important, which meant nothing felt important.

Copilot AI-assisted research synthesis


With the interview data collected, I needed to move fast from findings to actionable framing. I fed raw user quotes into Copilot to form HMW (How Might We) questions and establish hypotheses.

Reframing the problem into opportunity

  1. How might we help users immediately understand what needs their attention today?

  2. How might we surface the platform's value without overwhelming users with data?

  3. How might we reduce dependency on support by embedding guidance into the experience?


  1. How might we help users immediately understand what needs their attention today?

  2. How might we surface the platform's value without overwhelming users with data?

  3. How might we reduce dependency on support by embedding guidance into the experience?


Established hypothesis

If we redesign the dashboard to be more engaging and surface relevant, actionable information in one place, then users will better understand the platform's value, take actions more quickly, and engage more frequently with key features.


Established hypothesis

If we redesign the dashboard to be more engaging and surface relevant, actionable information in one place, then users will better understand the platform's value, take actions more quickly, and engage more frequently with key features.


design phase

design phase

AI generated the options.
I chose the direction.


Two months to ship meant I couldn't afford slow exploration. I used AI tools to diverge fast in week one and commit by week two. Research was clear: users needed to see what required attention immediately. We identified six modules already buried across the platform - appointment trends, critical notifications, recently visited sections, capacity planning, training centre, and persistent help - and surfaced them all on the home screen.


AI generated the options. I chose the direction.


Two months to ship meant I couldn't afford slow exploration. I used AI tools to diverge fast in week one and commit by week two. Research was clear: users needed to see what required attention immediately. We identified six modules already buried across the platform - appointment trends, critical notifications, recently visited sections, capacity planning, training centre, and persistent help - and surfaced them all on the home screen.

AI generated the options.
I chose the direction.


Two months to ship meant I couldn't afford slow exploration. I used AI tools to diverge fast in week one and commit by week two. Research was clear: users needed to see what required attention immediately. We identified six modules already buried across the platform - appointment trends, critical notifications, recently visited sections, capacity planning, training centre, and persistent help - and surfaced them all on the home screen.

AI generated the options.
I chose the direction.


Two months to ship meant I couldn't afford slow exploration. I used AI tools to diverge fast in week one and commit by week two. Research was clear: users needed to see what required attention immediately. We identified six modules already buried across the platform - appointment trends, critical notifications, recently visited sections, capacity planning, training centre, and persistent help - and surfaced them all on the home screen.


Text-to-layout exploration

I described concepts in plain language, and Figma Make generated 3 structural variations, starting points for reaction, not finished designs. The team selected notifications-first. From there, the work was manual craft.


Three mid-fidelity wireframes generated by Figma Make to validate


Text-to-layout exploration


I described concepts in plain language, and Figma Make generated 3–4 structural variations — starting points for reaction, not finished designs. The team selected notifications-first. From there, the work was manual craft.


Three mid-fidelity wireframes generated by Figma Make to validate


Text-to-layout exploration

I described concepts in plain language, and Figma Make generated 3 structural variations, starting points for reaction, not finished designs. The team selected notifications-first. From there, the work was manual craft.


Three mid-fidelity wireframes generated by Figma Make to validate


Text-to-layout exploration


I described concepts in plain language, and Figma Make generated 3–4 structural variations — starting points for reaction, not finished designs. The team selected notifications-first. From there, the work was manual craft.


Three mid-fidelity wireframes generated by Figma Make to validate


Notifications as the front door

Buried alerts became the dashboard's primary view - urgent items top, informational below. The homepage went from a wall of everything to a view of what matters now.


Before

After

Capacity planning that shows the full picture

A real-time view of clinician availability within the capacity planning module, showing who's on shift, who's absent, and where gaps need covering. Previously, practice managers pieced this together from separate systems. Now it's one glance at the dashboard.


Before

After

Modular data cards

Bringing key KPIs directly onto the dashboard changed everything. Dense tables became modular cards, giving users the numbers that matter without clicking into reports. Scannable in seconds.


Inline guidance over documentation

Hidden training resources replaced with contextual help within the workflow. Guided tours tested poorly; contextual hints at the point of need had significantly higher engagement.


Key design decision and trade-off

Stakeholders pushed back - "It feels too busy." Every team had added "just one more metric" over time. We stripped it back, chose clarity over richness, and tested with users. The response confirmed the direction. That became the phase one build.

Key design decision and trade-off

Stakeholders pushed back - "It feels too busy." Every team had added "just one more metric" over time. We stripped it back, chose clarity over richness, and tested with users. The response confirmed the direction. That became the phase one build.

Vibe-coding time - From static wireframes to working phase one dashboard prototype

With the phase one direction validated, I vibe-coded an interactive mid-fidelity prototype of the full dashboard, real interactions, navigation, and every new feature behaving as intended. Stakeholders could react to it, not just review static screens. Developers could inspect interaction logic and edge cases before writing production code, cutting handoff ambiguity.


Three mid-fidelity wireframes generated by Figma Make to validate


Vibe-coding time - From static wireframes to working phase one dashboard prototype

With the phase one direction validated, I vibe-coded an interactive mid-fidelity prototype of the full dashboard, real interactions, navigation, and every new feature behaving as intended. Stakeholders could react to it, not just review static screens. Developers could inspect interaction logic and edge cases before writing production code, cutting handoff ambiguity.


Three mid-fidelity wireframes generated by Figma Make to validate


delivery phase

delivery phase

What we shipped in phase one

A redesigned dashboard with new Optum brand UI. Practice Insights module with appointment activity, DNAs, utilisation, and mode of contact, each with trend indicators. Critical notifications panel with colour-coded alerts. Capacity planning widget with weekly slots and progress rings. Training Centre and a persistent help panel.


What we shipped in phase one

A redesigned dashboard with new Optum brand UI. Practice Insights module with appointment activity, DNAs, utilisation, and mode of contact, each with trend indicators. Critical notifications panel with colour-coded alerts. Capacity planning widget with weekly slots and progress rings. Training Centre and a persistent help panel.


What we shipped in phase one


A redesigned dashboard with new Optum brand UI. Practice Insights module with appointment activity, DNAs, utilisation, and mode of contact, each with trend indicators. Critical notifications panel with colour-coded alerts. Capacity planning widget with weekly slots and progress rings. Training Centre and a persistent help panel.

New dashboard, before and after (Left to right)

New dashboard, before and after (top to bottom)

Practice insights would display the most recent information on the dashboard.

Practice insign would display
most recent info in the dashboard

Capacity planning, including team availability.

Practice insign would display
most recent info in the dashboard

Notifications are now accessible from anywhere, either through the dropdown or on the page.

Practice insign would display
most recent info in the dashboard

Notifications are now accessible from anywhere, either
through the dropdown or on the page.

Notifications are now accessible from anywhere,
either through the dropdown or on the page.

Upcoming and past training sessions are now presented to users, who can add
events to their calendar or book a free 1-to-1 call with support.

Practice insign would display
most recent info in the dashboard

impact

What changed after phase one launched:

We plugged Pendo into the APEX platform and are monitoring changes. Three months after phase one launched, the numbers reflected what we had heard in testing.

What changed after phase one launched:

We plugged Pendo into the APEX platform and are monitoring changes. Three months after phase one launched, the numbers reflected what we had heard in testing.

What changed after phase one launched:

We plugged Pendo into the APEX platform and are monitoring changes. Three months after phase one launched, the numbers reflected what we had heard in testing.

IMPACT AFTER PHASE ONE

IMPACT AFTER PHASE ONE


70%

Reduced navigation drop-off and

improved feature discovery

18 points

User satisfaction scores improved

38%

Orienting faster and acting

with more confidence

63%

Backlog reduction

63%

Backlog reduction

47%

Increased employee productivity

47%

Increased employee productivity

Our hypothesis was validated: by surfacing relevant, actionable information in one place, users understood the platform's value better, took actions more quickly, and engaged more frequently with key features.

Support tickets dropped 23%. Critical alerts were no longer missed. The platform stopped being a barrier.

Our hypothesis was validated: by surfacing relevant, actionable information in one place, users understood the platform's value better, took actions more quickly, and engaged more frequently with key features.

Support tickets dropped 23%. Critical alerts were no longer missed. The platform stopped being a barrier.

Reflection

The hardest part wasn't designing features; it was deciding what not to show. Reducing cognitive load changed how users engage. AI tools let me explore 3× more directions in a short time and surface framings I wouldn't have written myself. The results belong to the research, the testing, and the design decisions. AI changed the path to the answer - not the answer itself.

What followed

Engagement continued for 10 months post-launch. Built additional features and developed the web platform.


but wait

There is more projects to check

There is more
projects to check

MOde case study

Mode


Turning a crypto wallet app into a product people trust and come back to


Redesigning the crypto app increased in-app transactions by 40% and improved the App Store rating from 2.2 to 4.1.

B2C

Crypto

Blockchain

Read case study

TrialTrucker case study

Ixico / TrialTrucker


New upload process for large scan files


A new handling process for large MRI, CT, and PET files increased productivity by 47% and stopped delays.

B2B

SaaS

HealthTech

Read case study

If you’ve made it this far, we should probably talk …


I’d love to help you solve your

next challenge.

If you get that far …


Let's work together to find solutions for your business challenges.

let's talk

let's talk

Email

Linkedin

London — Berlin — New York & sunny places

© 2026 Created and build by rado

TOP

Email

Linkedin