What if the recommendation actually listened to the assessment?

This is a small multi-page version of the same idea, built the way a real assessment tool would be: separate pages for each step, server-side session state carried between them, and a recommendation engine that runs on the server — not a single page that recalculates in the browser.

Choose a mode before you start. Legacy behaviour reproduces what the current tool actually returns. Evidence-based runs the engine that reads your answers.