Improved Information Architecture
Principal UX Designer at Capital One
The Problem
The signal came from a single usability session with eye tracking. A customer trying to find the account settings feature gave up after a full minute of searching, and it became that this wasn't a one-off. The information architecture itself was working against customers, particularly those who held products across multiple lines of business.
The deeper issue was structural. Consumer Bank, Consumer Card, and Auto had been designed in parallel, with each team optimizing for its own customer in isolation. Customers who held products across more than one line, which was a meaningful and growing segment, were navigating an experience that hadn’t been designed for them.
The Approach
I led the Consumer Bank IA research effort and built it as a cross-team initiative.
Multiple navigation models, tested head-to-head. Rather than iterating on the existing IA, I developed a set of alternative navigation models with different organizing logics. I ran a series of tree tests in Optimal Workshop with existing Consumer Bank customers, then benchmarked each model against the current navigation as a control. The tree-test format meant we could measure findability and task completion at scale, without the visual design influencing results.
Cross-team partnership with Card and Auto. I worked directly with UX designers and researchers in Consumer Card and Auto teams to design navigation models that worked for customers holding products across multiple lines, not just within one. Each team had context the others didn't, about customer behavior, account structure, and product-specific constraints, and pulling that into a shared IA was the only way the result was going to hold up across journeys.
Findings translated into stakeholder-ready reports. Tree-test data is only as valuable as the decisions it changes. I built findings reports that gave stakeholders a clear read on which model performed best and why, so the recommendation could move into design and build without re-litigating the research.
The Impact
The winning navigation model significantly outperformed the existing structure on the metrics that mattered most:
98% task success rate on the recommended model, vs. 74% on the control. A 24-point lift in customers' ability to find what they were looking for.
Notably faster task completion times. Participants reached their destination meaningfully quicker on the new model, reducing the friction that had driven the original abandonment behavior.
A shared IA foundation across Consumer Bank, Card, and Auto. The cross-team partnership produced navigation logic that worked for multi-product customers, closing a gap that previous siloed work hadn't addressed.
What I Took Away
A few things from this project I bring to every role:
One usability session can surface a structural problem. The participant who abandoned the search wasn't an outlier, they were a signal. Treating individual moments of friction as data, not anecdote, is what made the case for the broader IA work.
Cross-team research outperforms cross-team alignment. Pulling Card and Auto into the research itself, not just looping them in at readout, produced a model that actually worked for shared customers. Alignment after the fact rarely does.
Test multiple models, not just one improvement. Comparing several navigation logics against the control made the recommendation defensible in a way that a single redesign never would have been.