A platform built for data that nobody could actually use
John Deere's MAI (Manufacturing, Analytics and Insights) is a SaaS enterprise web application to help field managers across different manufacturing levels track relevant analytics, Key Performance Indicators, and derive insights to make data-driven business decisions. When I joined the project, the platform had a fundamental problem: it was built around the database structure, not around how field managers actually think about their work.
"The legacy system was organized to reflect how the data was stored - not how field managers thought about their responsibilities."
Field managers faced a challenge where their data access lacked personalization - they saw everything regardless of relevance to their role, creating significant cognitive load.
Information was spread across various tables and databases in a way that reflected database architecture, not user mental models - making it difficult to retrieve relevant data efficiently.
Static reports and outdated dashboards meant managers were making decisions on data that could be hours or days old - in a manufacturing context where real-time accuracy matters.
Understanding the organizational hierarchy
I conducted stakeholder interviews and contextual inquiry sessions with field managers across different levels of the manufacturing hierarchy - Plant managers, Group supervisors, Workcenter leads, and Material coordinators. Each role had fundamentally different data needs, yet the legacy system served them all the same view.
Rebuilding the sitemap around the org hierarchy
The core design decision was to restructure the system architecture around role-based access - breaking down the existing database table structure and replacing it with an information architecture that reflected the organizational process hierarchy: Plant → Group → Workcenter → Material.
This hierarchy formed the foundation of the application's structure, ensuring that users could navigate the system intuitively based on how they understood their own organization. Each role received a personalized dashboard showing only the data relevant to their responsibilities - with the option to drill down or export for deeper analysis.
I designed a dual-view system - graphical representations for trend analysis and KPI monitoring, tabular views for precision data extraction about specific products. This gives managers control over how they see the data without forcing a single representation on everyone.
Each dashboard module reloads separately at regular intervals, ensuring that managers see up-to-date information on each section with automated data pulls - eliminating the need to manually refresh the entire page to get current numbers.
Wireframes and design system in parallel
I developed hi-fidelity wireframes alongside a bespoke design system to ensure the delivery team had everything needed to implement accurately - not just beautiful screens, but the underlying component library and design tokens that made those screens reproducible.
Production-ready screens
The final UI mockup designs aligned seamlessly with the John Deere brand identity and catered to the specific use-case requirements. Screens included: role-based landing pages, parent dashboards, configuration settings, and data visualization options.
Measurable improvements across all success metrics
User satisfaction showed a significant increase following implementation. The new system's Net Promoter Score increased substantially, indicating greater user advocacy. The −50% reduction in dashboard generation time gave field managers quicker access to critical analytics and KPIs for time-sensitive manufacturing decisions. The +45% improvement in data-driven decision-making reflected managers' ability to track relevant metrics more effectively and identify trends that were previously invisible in the legacy system's disorganized tables.