Figma is where I do most of my work, but it is rarely where I start. The fidelity it gives you for free is exactly the problem early on: a rough idea looks finished before it has earned the right to.
When something looks polished, people respond to the polish. Sketches do not have that problem. A few boxes and arrows on paper invite “what if it worked completely differently” in a way a neat artboard quietly discourages.
Cheap ideas are honest ideas
A sketch costs ninety seconds, so throwing it away costs nothing. That changes how I work: I will explore five structures on paper before I would have committed to one on screen. The bad ones get discovered early, where being wrong is free.
Figma for the questions paper cannot answer
Once the structure feels right, the screen earns its place: real spacing, real content, real states, the things you cannot fake with a pen. By then I am refining a decision rather than making it, which is exactly what that level of fidelity is good for.
It is less a rule than a sequence. Think in the cheapest medium that can answer the question in front of you, and only move up when you have to.