I structured the Concord research across three methods. I ran JTBD sessions with contractors, sellers, and admins to map what each role actually needed from the platform — the findings revealed that the core job was not tracking orders but maintaining confidence that nothing had slipped between handoffs. A heuristic evaluation against the existing platform surfaced critical failures: delivery and load status only appeared after user input with no immediate feedback, terminology like "Projected" and "Compare" didn't match how field teams thought about their work, and users had no clear way to reverse decisions once made. I then designed a revised information architecture and ran a tree test across the restructured navigation. Task success on the existing structure averaged 51%, with the steepest failures concentrated in the invoicing and order tracking sections. The redesigned IA brought that to 86%, a 35 point improvement I confirmed before redesigning a single screen.