Embeddable Assessment Widget

A retailer needed a guided product-discovery flow that could ship as a standalone campaign page, embed inside an existing site, or pop as a modal — without maintaining three separate codebases.

The recommendation logic itself wasn't trivial: an eight-axis questionnaire mapped to weighted product cohorts, with the result needing to feel curated, not algorithmic.

One Next.js codebase, three render modes selected by a single flag. State machine for the questionnaire so navigation, validation, and result resolution are all explicit transitions.

Recommendation engine kept in plain TypeScript: each axis has a weighting matrix; the result is the highest-scoring cohort, with tiebreakers based on context.

Locked the type ramp and the questionnaire pacing so the widget reads as editorial copy, not a form.

Live across multiple deployment surfaces. Same widget runs as the campaign landing, embedded inside the main site, and as a modal launched from product pages.