Modernized a 16-Year-Old Legacy Desktop App for SaaS
Rebuilding H5's internal, VPN-only DART tool into a browser-based SaaS product on the Relativity platform.
ROLE
Lead UX Designer and Researcher
15 yrs
of design debt resolved, workflows streamlined, and a design system introduced
COMPANY
H5 / Lighthouse
TIMELINE
11 / 2020 - 1 / 2023
100%
task completion rate in final usability round, up from 80%
TEAM
1 report, 3 Product Managers, 50 Engineers
Desktop→SaaS
rebuilt from a VPN-only tool onto the Relativity cloud platform
The Problem
A 2005 desktop tool, still running in 2020.
DART was H5's internal desktop application for search strategists and reviewers, collection creation, complex query building, document analysis, accessible only over VPN on a virtual machine. Sixteen years without a UX designer had left real design debt, and the business needed it rebuilt as a browser-based product it could host on Relativity, legal-tech's version of Salesforce, to reach customers beyond internal VPN users.
Role & Constraints
PARTNERED ON
Product management and engineering throughout; a new Lighthouse design team post-acquisition.
CONSTRAINTS
VPN-only legacy access, sixteen years of accumulated debt, a design system alignment and expansion mid-project.
OWNED
UX design for DART's transformation into a SaaS product, managing a mid-level designer while supporting Matter Analytics in parallel.
WHAT CHANGED MID-FLIGHT
The PM departed and H5 was acquired by Lighthouse, both landed partway through the build.
The Approach
4 steps, one vision.
1. DISCOVERY: Usability audit against Nielsen's heuristics, plus user research into delights, pain points, and workarounds.
2. DEFINITION: Scoped an MVP against user, business, and design/technical constraints.
3. DESIGN & ITERATION: Benchmarked Salesforce and Relativity, then tested the Formula Builder through two R.I.T.E. rounds.
4. DELIVERY: Shipped incrementally, realigning to Lighthouse's design system mid-migration.
Key Decisions
Round 1: 80% success
Cut and consolidated the Formula Builder's buttons
Why: Users confused similarly-named buttons, mistook the move icon for a menu icon, and read red icons as errors.
Result: Removed color-as-signal outside of CTAs, consolidated redundant buttons, made the remaining ones more prominent.
Round 2: 100% success
Fixed the smaller usability gaps round one surfaced
Why: Some users still struggled with group creation, unclear micro-tasks, and small tap targets.
Result: Disabled-state for uneditable fields, larger tap targets, buttons de-emphasized to match the new design system, every tester completed every task unaided.
The Work
2005: Legacy desktop app
VPN-only, virtual-machine desktop interface.
2023: Streamlined SaaS app
Browser-based, cloud-hosted redesign for the Relativity platform.
Memorable Quotes
“This is way better than what we’ve used for the last 10 years!”
Outcome
What shipped.
DART now runs on the Relativity platform, reaching customers beyond internal VPN users, with simplified workflows, greater feature discoverability, and reduced time-to-task. H5 was acquired by Lighthouse during this window, the modernized platform was part of the case for it.
Reflections
What I would do differently.
Clickable prototypes won arguments faster than static comps ever did. If I ran it again, I'd get in front of real users sooner, I spent more time than I'd like learning the problem secondhand through the PM before user conversations started.