archive

Instead of asking how to make something succeed, ask what would guarantee its failure — then design those things out. For a clinical workflow: list everything that would make a nurse abandon the screen, and remove those obstacles before adding anything new.

Don’t remove a rule, a step, or a piece of UI until you understand why it was put there. Legacy healthcare workflows are full of fences — some guard real safety constraints, some are just scar tissue. Find out which kind you’re looking at before you simplify.

People don’t buy products; they hire them to make progress in a specific circumstance. The unit of analysis is the job, not the persona. When a clinician falls back to a phone call instead of the portal, ask what job the phone call is being hired for.

Two cycles of diverge-then-converge: discover the right problem, define it, then develop options and deliver the right solution. Its real value is permission — to explore widely before committing, twice, instead of converging on the first plausible answer.

Clinicians are interrupted every few minutes. Every screen should assume it will be abandoned mid-task: save state aggressively, make re-entry obvious, and never punish someone for leaving. The happy path in healthcare software is the resumed path.

Most people never change a default. Whatever ships pre-selected is, in practice, the product decision — the settings page is mostly where a team’s indecision goes to live. Choose defaults as carefully as you’d choose for someone who will never see the options.