Scentwise
Overview
Side project since 2025 — fragrance discovery on Next.js and Supabase. Still early; figuring out what features earn their place and what gets cut.
TL;DR
- Concept: a fragrance discovery platform — search, explore, and understand scents without the usual retail clutter.
- Status: in progress since August 2025. MVP-first — ship small, learn fast.
- Stack: Next.js, Supabase, TypeScript, deployed on a modern serverless pipeline.
- Audience: fragrance lovers who want better tools, and curious newcomers who need a gentler on-ramp.
- Current focus: core data model, search, and a UI that feels calm — not another loud e-commerce clone.
The idea
Fragrance discovery is broken in a specific way: too much marketing language, too little useful context, and almost no tooling that treats scent exploration like a craft rather than a checkout funnel. Scentwise is my attempt to fix the slice I care about — helping someone answer "what might I like?" and "why?" without drowning in brand copy.
What I am building
The first version is deliberately narrow. I am not trying to be a full retailer or a social network on day one. The MVP is about discovery mechanics: structured fragrance data, search that respects how people actually talk about scent, and pages that explain a fragrance in plain language.
- Fragrance catalog with consistent metadata — notes, families, and context that survives a search.
- Discovery flows tuned for browsing, not just filtering by brand.
- A fast, readable UI — mobile-first, minimal chrome, no dark patterns.
- Auth and user data via Supabase — lightweight accounts when they earn their keep.
Stack & approach
Next.js for the app shell and rendering strategy, Supabase for database and auth, TypeScript throughout. The architecture bias is boring on purpose: fewer moving parts, easier to iterate while the product shape is still fluid.
Where it is now
Active development. The product surface is still shifting — some features will ship, others will get cut once real usage proves they do not matter. I would rather publish an honest in-progress case study than pretend the scope is frozen.
What is next
- Tighten the core search and detail experience before adding social or recommendation layers.
- Validate the data model against real catalog imports — edge cases always show up late.
- Ship a usable MVP, then decide what "version two" actually needs based on feedback, not assumptions.