SPSS Modeler pioneered visual machine learning: since the 1990s, analysts have built predictive models by connecting nodes on a canvas instead of writing code. As senior design lead, I modernized that canvas for today's model builders, making large, complex flows faster to build, easier to read, and more accessible, without breaking the workflows long-time users depend on.
Client
IBM
DELIVERABLES
CANVAS UX SYSTEMS DESIGN ACCESSIBILITY BRANDING
Year
2022
Role
SENIOR DESIGN LEAD
Design strategy
SPSS Modeler had a loyal base of analysts who have built models on its canvas for decades, and a new generation of model builders used to modern, fluid tools. Redesigning for the newcomers risked alienating the experts; designing only for the experts meant losing the next generation. My strategy was modernization without disruption. The core mental model of connecting nodes on a canvas stayed intact, so long-time workflows kept working. What changed was everything around it: readability at scale, responsive direct manipulation, and accessibility. Experts felt the product get faster, and new users found a canvas that felt current from the first session.
Canvas research
Readable at scale. Smart auto-layout and semantic grouping with collapsible flows keep large models understandable. Direct manipulation that responds. Clearer drag-and-drop feedback, contextual zoom and pan, and reliable undo and redo make building faster and less error-prone. Accessible by default. Better touch and keyboard support opened the canvas to more users.
