0 1 - I N T R O
Building trust in the shadow of a failed product
02 - Quick Stats
1,000+
7% to 18%
Response rate improvement
55
Clients actively using the product
9 days
of development work saved per hub
03 - The Problem
A project with baggage
For job seekers
Branded employer hubs embedded within career centre sites showcasing company culture, values, and career opportunities. The goal to give job seekers the context they need to apply without leaving the platform.
For clients
A self-service page builder in the CMS. Clients assemble hubs from content sections including text, images, and branded containers, which they self publish to their career centre. No developer or design resource required.
04 — Research Insights
Users wanted design freedom they weren't trained for
Through research with clients, marketing teams, CSMs and site owners, I found the tension that defined the project. Users wanted full creative control but lacked the expertise to use it well. Only 1 in 6 understood basic accessibility.
What users wanted
Unlimited strucutral control
Freeform colour options
Pixel-perfect customisation.
Designs that match employer's career site exactly
What users needed
Smart constraints
Accessibility guardrails
Design-vetted templates
Forgiving, responsive layouts
Research evidence
05 — Critical decisions
The beta test that changed the project
The PM wanted full colour customisation across the product and the developers had reusable code ready to go. I was new to this team, and pushing against more user control sounded like repeating the mistake that killed v1.
My compromise: Limit colour customisation to one section, test with beta users, and measure the outcome. Across six users over one month, five of six created inaccessible sites with the tool we offered.
The evidence proved our users needed controlled guidance.
Beta results (Before)
Final design (After)
My alternative: Section Styles. Three preset text and background combinations built from design system brand tokens, all WCAG 4.5:1 compliant.
Plain white, neutral grey, and a branded container.
Users still got visual variety across their hub, but every option was accessible by default. Simple choices, professional outcomes.
The outcome
Rolling back the beta required a late-night maintenance session to republish all sites with defaults. This would be very difficult now at 1,000+ live hubs. But it built lasting trust in my design judgment, and Section Styles became the standard across the platform.
06 — Design Approach
The pattern scaled across the product
That approach — advocate, compromise, gather evidence, earn trust — became my playbook.
Image resolution
The team wanted to use native resolution for image uploads. It was the easiest way to implement the feature. Based on our initial research I was concerned about our users' lack of understanding on appropriate image resolution.
I advocated for fixed image frames with transparent backgrounds and fill/zoom controls. This gave us occasional grainy images, but consistent and responsive layouts.
Native resolution (Before)
Fixed frame (After)
Pixel padding
A client requested that we give them pixel-perfect padding control for every part of their site.
This time I was able to push back and advocate for further investigation and an alternative approach.
The client was having difficulty organising long text content. So insted we provided drop in page dividers
Padding controls (Before)
Page dividers (After)
07 — User Testing
Building team trust with inclusive testing
I invited the entire sprint team to observe user-testing and take part in an insights and solutions workshop.
The team found it highly rewarding to see their hard work perfoming well with users and had plenty of great insights, and ideas to improve the product.
We tested with 6 users of varying design experience
Research evidence
Positives
Improvements made
07 - The Outcome
1,000+ live hubs
Employer hubs built across 55 clients. Significantly better adoption than its predecessor
From 7% to 18% response rate
More than double the application response rates for jobs advertised on employer hubs
80+ hubs created by individuals
Showing a strong product-market fit, and ease of use at scale
9 work days saved per hub
When compared to the manual sites requirement, built internally. A huge operational saving
"It's all easy enough to use. It makes sense and it's very user-friendly."
Abbie Boyd, Operations Specialist, Pharmiweb
"The difference between the two is night and day. It's a lot easier on employer hub."
Kevin Kirchmyer, Sales Operations, Science Careers
The final product
Manage hubs on a dashboard, edit pages, design content in an editor, preview in browser, publish directly to the live site.
Smart constraints over unlimited control
Pre-built components with limited customisation
Responsive by default
Layouts adapt automatically across devices
Accessibility guardrails
Brand colour prests, readable contrast, image ALT text, guaranteed compliance
No design knowledge required
Marketing teams and CSMs can build professional sites with ease
Scalable and adaptable CMS
A design system of menus and components that scale, across a variety of clients, whilst maintaining usability
08 — On Reflection
What I learned
Testing builds credibility
If I had the confidence I had today, I may have push harder against colour customisation. But the beta approach built my credibility through shared evidence and my own confidence in the design process. The team saw the problems themselves, not just heard me predict them. This created lasting trust that served the entire project.
Beta feeback is noisy
Opening up a beta generates a lot of input, but users' prescribed solutions aren't always the real need. Two beta users asked us to extend colour control further, which would have made the problem worse. I'll always question whether feedback fits the design ethos before acting on it. Our job is to solve the underlying need, not implement the suggested fix.
What's next?
Career paths are the most valued feature, but they need data volume to be trustworthy. Smaller clients often fall below our data confidence threshold and clients retain ownership of their data, so we can't aggregate across them. The data science team is exploring supplementary sources like O*NET to close this gap.
We also found that broadly generic roles, such as scientist or university professor, often return low-value insights because their skills depend more on industry than job title. We're refining our taxonomy to separate general competencies from areas of speciality.
As a result, we haven't progressed to the personalisation phase of the product. It's a missed opportunity, but it validates the our choice of a phased approach. Better to discover data limitations at phase 2 than to ship personalised recommendations on unreliable foundations.
