In order to understand how the product works, we inevitably had to dig deeper into the domain—in this case, the heart. It was both fascinating and surprising to see that even in cardiac electrophysiology statistics there are normally distributed phenomena. :)

During the discovery phase, we uncovered the existing solution, whose assembly required 15 years and the efforts of many people. We found hundreds of screens and features whose design rationales we couldn’t always reverse-engineer in months. Through task analysis and targeted shadowing, we managed to grasp end users’ mindsets, pain points, and what would truly make a difference for them.

These “star destroyers” always arise organically: users signal their needs, and development teams implement them to the best of their knowledge. That’s why maintaining consistency across these interfaces is such a challenge. With today’s thinking, it’s clear that a design system offers a solution to this problem - indeed, that’s partly why they came into being. We built our design system from scratch, paying attention from day one to scalability, high component variability, and minimizing custom elements.

For software as a medical device, there is only limited room for purely aesthetic considerations in the visual strategy - the high contrast ratio, the interface’s strong risk-mitigation capabilities, and the clarity of tasks and feedback are not just important but critical. These priorities guided our value choices as well.

As is typical in medical-device development, we cannot continuously consult with physicians and healthcare professionals in the field. This gap is bridged by ongoing collaboration and validation with domain subject-matter experts. That’s exactly what we did: we presented and discussed our ideas and prototypes at various stages with experts, which in turn yielded a wealth of strategic insights for our design.

Prototyping is an exotic challenge for a product of this size and depth, so we applied a mixed strategy—for example, we only tested interactions requiring highly specialized expertise on hard-to-reach participants, while more straightforward, intuitive tasks or higher-level navigation flows were validated with users who had even a basic familiarity with the domain.