SKU PROJ-004 · Release 2024
At CEA, a research team simulating supply chains with machine learning, without a shared language. The digital twin's design system: simple, adopted down to the least experienced profiles.
An R&D project: using machine learning to simulate real-world behaviour, starting with supply chains. The team was very mixed: two designers, developers, the CTO. With no shared design foundations, every screen reinvented its components and delivery was getting stuck. The challenge: add structure without freezing things, with a system that serves a research team.
Ideation workshops and retrospectives clarified the vision of the digital twin. We identified the needs and defined design principles (“What values do we stand for?”, “How should components be used?”), to align everyone, from the designers to the CTO.
The system rests on an atomic approach: PascalCase names, a six-step size scale, and a detailed inventory of components and variables to map the redesign and prioritize. A single requirement: simplicity, for a system anyone understands, even without experience.
The style guide was built atom by atom (variables, icons, type, spacing). Every component ships with its documentation and release notes. Properties avoid useless variants, and contribution guidelines keep quality up over time. Regular retrospectives tracked what matters in R&D: team satisfaction.
The design system structured the team's work and was adopted, including by the least experienced profiles, which the retrospectives confirmed.