Data showed that users were experiencing major usability issues across the entire app. Since ROSS had gone through many changes over the past few years without an established design system, there were visible traces of individual styles that varied from one designer to another causing usability issues and inefficient release cycles. Moreover, user research was limited and inconsistent and didn't play a role in product decisions.
Initially ROSS had acquired customers from large firms looking to offer their employees access to multiple legal research tools and relying on ROSS as the fastest tool for finding relevant caselaw, but as the product and its competitors evolved, ROSS shifted focus on solo practitioners and small firms looking to reduce time spent on legal research in order to grow their practice. ROSS also offered free access to faculty and law students practicing their legal research skills.
When I first joined ROSS in the end of Q2 2019, the company was preparing for Q3 sprint planning, my first priority was to help identify design requirements for an app redesign. I organized and facilitated cross-functional ideation workshops and design studios to create alignment on a vision for a redesigned app that solved the most critical user problems, defined the design operations and helped estimate and plan the design sprints.
In the past, research at ROSS had been inconsistent and limited. I created a research base where we continuously collected raw qualitative data from sales calls, surveys, user interview and usability testing sessions. I synthesized and organized our backlog in airtable making it available for the entire team to contribute and consume to research. I held a weekly grooming session and planned appropriate research activities to support ongoing design stories, and I created templates and frameworks to help the design team recruit, plan studies and conduct user testing on their own. In addition, we held monthly usability reviews on fullstory and created design tickets to solve small usability issues.
I defined ROSS’s design operations based on transparency, efficiency, and collaboration. Our workflow included design sprint planning, user research, user flows, collaborative design studios, design reviews, wireframing, high fidelity mockups, and user testing. I worked with the product team to create design tickets in airtable and to estimate efforts and pair with designers where they needed support, held daily design huddles and weekly design critiques, UX briefs and component documentation, best practices research and book clubs relevant to ongoing design tickets.
I directed a phased approach to implement our design system by first documenting and implementing reusable components in each feature release and next, I conducted a brand perception survey and a desirability test to identify gaps in our visual design and used that to define guiding principles for ROSS's visual identity. I onboarded a visual designer to create our typography, iconography, spot illustrations, and logo design, and I created our component library in Figma and collaborated with front-end engineering to gradually implement the design system.
While I oversaw our operations and the redesign of multiple features by our design team, I operated as a hands on designer throughout the entire design cycle from discovery and user research, product requirement gathering and prioritization, userflows, wireframes, high fidelity mockups and user testing on multiple high impact projects including the redesign of our NLP search, jurisdiction filters, statutes and regulations search and topic modelling.
Below are screenshots of the main areas of the app before and after the UX revamp and the branding update and implementation of ROSS's design system.
ROSS's core feature, the AI search was difficult to use and didn't match users' expectations and existing mental models. The updated search and result page included revised navigation, improved information architecture for displaying results and a new interaction model for an omnibar with query intent detection, contextual search tips, recent searches, named entity recognition (NER) for automatic filters and improved groupings of advanced features and filters.
User research showed that most users were frustrated by the lack of control over jurisdiction filters. They wanted to narrow their results by specific courts, but our UI only displayed regions and limited federal courts. The granular court data existed in our system, but the taxonomies didn't support faceted groupings of courts. After a series of user testings and a tree test, I defined a new information architecture that allowed users to quickly search through state or federal tabs by typing the state name and focusing their search on specific court types and rank.
