The golden goose goes platinum.

All credit cards are not created equal, and there’s a mental calculation to be done on the value derived from each of them based on your circumstances. As soon as I arrived I identified significant interaction design challenges, and in my first week I proposed that alongside a cadence of quick-win experiments, I would also investigate a north star — a fundamentally better way to solve what I call the ‘million to one’ problem.

This wasn’t new territory. I was responsible for exactly this at Thomson Reuters a decade earlier — their flagship legal research tool Westlaw AU. Getting from a long list to a manageable short list to the thing you’re after, with the least friction possible. Legal precedents or low-fee credit cards — same mental model.

I hit the ground running identifying key areas to optimise: filtering, sorting, de-clutter, microcopy, interface uplift, and education. These were workshopped with stakeholders and prioritised through DVF.

The lead experiment targeted two hypotheses. First, the three CTAs were squeezed together and felt like part of the same action — de-clutter. Second, the word ‘apply’ creates click fear because users aren’t actually applying, they’re being transferred to the partner website — microcopy.

Using Optimizely, we tested a range of CTAs from the control ‘Apply’ (most click-fear) through to ‘View offer’ (most click-bait). The results were compelling but nuanced — the 106.73% uplift on ‘View offer’ was actually click-bait, not genuine intent. Users arrived at the partner site, realised there wasn’t an ‘offer’ as such, and left.

The winning variant was ‘Go to partner’ — the ‘Goldilocks’ moment that didn’t create fear or bait. We validated this with actual conversion-per-variation data, not just click-through. We then iterated by separating the compare action and relocating ‘more features’ to a more logical position under the feature bullet points — producing further significant double-digit improvements.

The existing sort and filter was broken. The zone that contained it didn’t make it clear which was a sort, which was a filter, and which affected estimated points earning. We were making the user think — hard.

Credit Card Selector was the golden goose, and any changes to how it worked were significant. We needed to prove beyond all reasonable doubt that this would be a must-win. I put together an exhaustive initiative brief highlighting the opportunities of following a ubiquitous pattern — backed by independent research.

The goal was clear: increase time on site through more interaction and immersion with the cards and their features. This would inevitably mean more conversions. But the attribution challenge was real — once users leave CCS, they’re on the partner website, and there’s only a 30% chance of linking the two sessions, especially if they’re not logged in.

I built a fully working, interactive HTML prototype using the same production JSON data for the credit cards. Not a Figma click-through — a real, working application that participants could interact with as if it was the new production website. I’ve been working with the open-source Westpac GEL framework since its inception, having been there at the start, and over ten years I’ve fine-tuned it into an exceptional prototyping tool — styled here in full Qantas livery.

That is why I am at my best when given a data-driven design challenge. I didn’t propose faceted search and wait for a dev team to tell me why it couldn’t be done — I built it myself, with real data, in a fraction of the time, to validate the problem to be solved.

The prototype included a feature I called ‘visual sort’ — inspired by Kayak’s ‘Best, Cheapest, Quickest’ pattern, where users can see the trade-offs at a glance rather than mentally parsing a table. I was a massive proponent of it, but I could read the room — stakeholders weren’t buying the departure from the existing table-based sort. So I compromised. Knowing when to push and when to land the bigger win is part of the job.

Moderated through the Askable platform, the testing thoroughly validated the problem and the proposed solution. The faceted search pattern aligned with users’ existing mental models from other marketplace experiences — they didn’t need to learn anything new.

That device-switching insight identified a gap: there was no way to move from one device to the other frictionlessly. This was prioritised as a backlog feature. We also used Dovetail AI to analyse the research findings, cutting synthesis time from days to hours.

Validated hypothesis. Detailed design. Full product plan. Convincing the execs it would have no unintended consequences. The faceted search went live towards end of 2025. Although hard numbers are under NDA, the uptick was significant — a hockey-stick graph that continues to climb.

The timing couldn’t have been better. The RBA’s recent decision on credit card surcharges means that if surcharges are banned or reduced, the value proposition of points-earning cards shifts — more consumers will be comparing cards on features and fees rather than defaulting to whatever their bank offers. The Credit Card Selector becomes even more critical as the primary decision-making tool.

As soon as this project completed, I was given the 'crown jewels' - leading design on the biggest changes to the Qantas Loyalty program in its 40-year history, re-imagining the My Account and Activity Statement pages that every single QFF member uses to track their progress within the program.

The ‘million to one’ problem isn’t unique to credit cards. I solved the same pattern for legal research at Thomson Reuters a decade earlier. This case study shows what thirty years of cross-domain experience looks like in practice: recognising the pattern, knowing the solution exists, building the evidence, and then building the thing — not handing it off, but actually building it. The prototype wasn’t a deliverable to be implemented. It was the proof.