01 | The Problem
A critical platform
heavily dependent on people's memory and experience.
CommBroker is where brokers find policy, submit applications, and manage compliance. Despite that weight, it had quietly become something brokers navigated by memory, bookmarks, and support calls, not by design.
What the heuristic review found
- Non-intuitive navigation that didn't match broker mental models
- Inconsistent content structure and templates across hundreds of pages
- Limited feedback and error prevention throughout key flows
- Poor accessibility compliance
- Large text blocks increasing cognitive load at critical moments
What the content audit revealed
- 36% of content streamlined through audit
- Significant content duplication and broken links throughout
- No reliable single source of truth for most topics
- Content structured around CommBank's internal teams, not broker tasks
- Brokers couldn't trust they were reading the right or latest version
The consequence
Brokers had compensated by building their own workarounds: bookmarks to pages they knew still worked, direct calls to CommBank for anything uncertain, peer networks to verify current policy. The platform had stopped being the source of truth. Brokers had become the source of truth, despite the platform's existence.
02 | Research & Discovery
Navigation was the problem.
Not the content.
The redesign was grounded in multi-layered discovery before any concept work began: desktop analysis, heuristic evaluation, content audit, Treejack IA testing, concept testing, and moderated usability testing.
Method 01
Content audit
Full content audit across 500+ pages surfaced duplication, outdated content, and broken links. Established the true scale of the problem before any design decisions were made.
Method 02
Treejack IA testing
Tested broker ability to locate information in the existing structure. Poor baseline success rates confirmed the primary failure was IA, not content quality. Navigation was the problem, not the words.
Method 03
Concept testing
Two concepts explored with brokers and broker assistants in moderated sessions. Discoverability, confidence, and language comprehension measured. Findings directly determined which direction went to delivery.
Research finding
Personas and jobs-to-be-done analysis showed brokers wanted simpler, more direct language, fewer interpretation steps between their question and the answer. They weren't struggling to understand CommBank's policies. They were struggling to find them.
03 | Concept Development & Testing
Two concepts.
One clear winner.
Two directions were developed and tested with real brokers. The decision wasn't made internally. It was made by the data.
Concept A
Familiar
Discoverability success
More familiar structure. Lower cognitive leap.
Concept B(CHOSEN)
Divergent
Discoverability success
Bold new IA concept. Higher task completion.
Concept B reduced navigation from 9 items to 7 by consolidating dispersed content into two hubs: Resource Centre (policies, processes, forms) and Loan Manager (applications and loans). Content was reorganised around broker intent, with search as the primary interaction model.
Home
Your Applications
Your Loans
Products
Credit Policy
Processes
Forms
Interest Rates
Training Hub
Home
Loan Manager
Resource Centre
Products · Credit Policy · Processes · Forms
Tools
Rates & Fees
Training Hub
Contact
Brokers described Concept B as "intuitive", "humanised", and "user-friendly". The contrast with the existing platform was immediate and unprompted. Brokers noted they could find what they needed without reaching for the phone.
"It feels like a guide-book."
The trade-off accepted
Some brokers noted initial unfamiliarity with Concept B's new structure, a short learning curve before efficiency gains kicked in. Testing showed this was an acceptable trade-off: brokers who understood the structure achieved dramatically higher success rates and reported significantly greater confidence. The data made the choice clear.
04 | Design Solution
Three pillars.
IA first, always.
This project reinforced that information architecture doesn't support the experience — it is the experience. Polish can't fix poor structure. Clarity can't be styled in. Three takeaways:
01
Information architecture as the primary UX lever
Rather than redesigning pages in isolation, the work prioritised a coherent IA above everything else: clear categorisation, removal of duplicate content, one source of truth per topic. IA decisions were validated with 219 brokers and assistants through iterative Treejack testing, refined based on success and directness metrics. Nothing was assumed. Everything was tested.
02
Centralised Resource Centre as single entry point
A single entry point for policies, processes, tools, and updates, with search as the primary interaction model. Brokers could find what they needed without navigating legacy silos or relying on institutional memory.
03
Content organised by broker intent, with multiple pathways
Content grouped by how brokers think about their work—by intent and task—rather than CommBank's internal structure. Multiple ways to find content according to their mental model, not a single prescribed navigation path.
05 | Reflection
Design operating
at a systems level.
CommBroker didn't just produce a flashy before-and-after screenshot. The most important work (content audit, IA validation, iterative Treejack testing, concept iteration with real brokers) happens before anything looks different.
The 98% discoverability figure matters. But what it represents matters more: brokers who could find what they needed without calling CommBank, without relying on a colleague's institutional knowledge, without a bookmark from three years ago that may or may not still work. That's not a UX win. That's a platform that finally does what it's supposed to do.
The foundation-first decision was the hardest to defend internally. It meant slower visible progress early on. But it's also why Gen-AI Instant Answers could be built on this platform rather than around it. Good design at a systems level creates options. That's what this work was for.
