Work · 05/2016 – 05/2022
Liberty Mutual Insurance
6 years leading information architecture (IA) and research for myLiberty — Liberty Mutual's enterprise intranet, used by 45,000+ employees across 8 global markets. The core problem: a decade of organic content sprawl had made the site nearly unusable. I led the IA redesign, research, and mega menu design that gave the platform a structure employees could navigate and content editors could maintain. The outcome was not just a new navigation bar, but a decision framework for where content belonged, how it should be named, and how hundreds of editors could maintain it consistently.
Context
A decade of content sprawl, and nobody could find anything
myLiberty had grown organically over 10 years: thousands of pages, hundreds of content editors, no shared structure, no placement rules, and no naming standards. Search had become the primary navigation mechanism, and it was absorbing failures the site structure should have prevented. Employee feedback about search was consistently negative.
The trigger was a SharePoint-to-Drupal platform migration. 400+ editors couldn't rebuild their content until a structural foundation existed: the IA had to be in place before any of them could start.
- 1,000+ pages, no consistent structure or placement rules
- Search became the primary navigation method, and barely worked
- 400 editors rebuilding independently = compounding fragmentation
- A validated site structure that reflected where employees expected to find information
- A content governance system 400 editors could apply consistently without ongoing UX oversight
- Navigation that made key resources findable without relying on search
- A structural blueprint scalable to international markets
- A repeatable decision framework, not a one-time fix
Research approach
Diagnosing the structural problems before designing the solution
Used a mixed-methods approach: IA-specific research methods to surface employee mental models and validate the proposed structure, layered with existing search analytics and support ticket data to diagnose where the current system was already failing employees.
Together, the methods covered different levels of evidence: card sorting shaped the categories, first-click testing tested labels and expectations, tree testing validated whether the structure worked without visual design, and moderated sessions explained why employees got stuck.
Content design
One page per topic, one place to maintain it
Every content page needed a structure: what it contained, how it was organized, who it was for. The central structural decision: manager-only content would live on the same page as all-employee content, not siloed in a separate section of the site.
This was a content governance decision as much as a UX decision: splitting topics by audience made the system harder to navigate, harder to maintain, and easier for content to drift out of sync.
A topic like "Benefits" existed in two places:
- Employee Center → all-employee benefits content
- Manager Center → manager-specific benefits guidance
Two pages to find, two pages to maintain. Managers who wanted the full picture — what their direct reports saw plus their own guidance — had to navigate to two separate sections of the site. Content editors updating the same topic in two places, with no structural relationship between them, meant content could drift out of sync or directly conflict.
One Benefits page, with role-aware content inline:
- All-employee content visible to everyone
- Manager content embedded in context, adjacent to the relevant all-employee content, not buried in a separate section
One page to find, one page to maintain. Managers see the full picture without navigating elsewhere. Content editors work within a single site structure.
Information architecture
The structural redesign
The redesign produced four top-level categories, each grounded in card sort data showing how employees naturally grouped content and tasks. Each category came with terminology standards and placement rules so 400 editors could rebuild consistently without needing UX sign-off on every decision.
Navigation design
Mega menu: making the full site structure as navigable as possible
The existing menu surfaced a curated subset of links chosen by content editors. It did not reflect the full site hierarchy, so employees could not use it to move systematically from a broad category to the resource they needed. The challenge was to expose enough structure to make the site navigable without turning the menu into an overwhelming sitemap.
Visual design: menu overlay height stays consistent regardless of how many subsections a section has.
Scalability: prevents future content bloat from degrading findability.
Impact
Outcomes of the IA redesign and mega menu
What this demonstrates
Systems thinking at organizational scale
| Capability | How it showed up at Liberty Mutual |
|---|---|
| Information architecture as systems design | The site structure was not just a navigation redesign. It came with terminology standards and placement rules that let 400+ editors rebuild and maintain content consistently without constant UX intervention. Navigation support tickets dropped from hourly to zero. |
| End-to-end system context | For six years, I was the primary UX and IA practitioner on myLiberty, with deep working knowledge of how the platform fit together, where it had failed before, and what its structural constraints were. That context shaped new features, market expansion, and content decisions across the platform. |
| Mixed-methods research to diagnose, not just validate | Card sorting revealed mental models; tree testing validated the proposed structure against those models; search logs and support tickets surfaced where the existing system was already failing. |
| Designing for scale and maintenance | The IA had to work at launch and scale to 45K across Liberty's global markets. The placement rules and terminology standards were designed to be maintainable by 400 content editors independently, with no ongoing UX oversight required. |
| Translating ground-level friction into platform decisions | Request a Service was a new section created to house tool and system links in one place, organized around what employees needed to do rather than what internal teams had named the tools. Plain-language task descriptions came directly from observing employees fail to identify the tools they needed. |
Full scope of work
6 years across every aspect of the platform
The IA redesign was the foundation everything else was built on. Over six years, as the platform grew — new features, new markets, new content initiatives, new accessibility requirements — the structural decisions made at the start shaped every one of them.
Related work
Self-guided cooking skill app: 7 interviews, 2 design rounds, review of 8 products
A home-decor brand run almost solo: a study in where AI helps and where humans decide
Voice-first Italian tutor with a modular AI pipeline and persistent learning state