a taste engine disguised as a shop
Scentum is not a conventional fragrance store. It is a tool for understanding what you like, why you like it, and what else might feel the same way. The AI is not a feature — it is the interface.
Taste as geometry
Most fragrance sites filter by note: cedar, rose, oud. Scentum goes deeper. Every perfume in our catalogue is mapped across five perceptual dimensions:
- Authenticity— how natural vs synthetic the materials feel
- Projection— intimate skin-scent vs fills the room
- Longevity— fleeting vs lasts all day
- Complexity— a single clear note vs deeply layered
- Versatility— very specific to one moment vs wears anywhere
These dimensions form a five-dimensional space. When you describe what you want — through search, through the guided Oracle, or through your reactions — we find the region of that space that matches your intent and surface the perfumes that live there.
Two perfumes can share every note and feel completely different. Two perfumes with nothing in common on paper can feel identical to wear. The geometry captures that.
The Oracle
The Oracle is a five-question guided flow. No typing required. You answer: what occasion, what mood, what memory or image, any constraints, and whether you want something familiar or surprising.
Each answer maps to a target in the taste space — no AI involved, just careful translation tables built from fragrance knowledge. Occasion shapes projection and longevity. Mood shapes authenticity and complexity. Novelty shifts how layered the result should be.
The Oracle resolves your answers into a search query and a vector target simultaneously, then hands both to the search engine.
Search, then refine
Free-form search understands taste language, not just keywords. Type “something quiet for a rainy morning” and the engine parses mood, occasion, and intensity — not just the words.
When results appear, a reaction bar lets you say: too much · warmer · love this. Each reaction rewrites the underlying query, keeping what worked and correcting what did not. Three reactions in a row can take you from a vague feeling to a very specific fragrance profile.
This refinement loop runs inside your current session. Nothing is stored between visits. There is no account, no history, no profile being built about you.
Similar perfumes
The similar perfumes on each product page are not based on marketing categories. They are calculated from three signals weighted together:
- Vibe alignment — shared aesthetic tags carry the most weight. A perfume that belongs in the same cultural register will feel like a true alternative.
- Accord family — shared dominant accords (woody, floral, smoky) contribute a secondary signal.
- Structural proximity — how close the two perfumes sit in the five-dimensional score space. Same projection, same complexity, similar longevity: they will wear like siblings even if their notes differ completely.
The community signal on each page shows how a perfume ranks against the full catalogue on each dimension — so you can see at a glance whether it is unusually quiet, unusually complex, or unusually versatile relative to everything else we carry.
The scent story
Every product page generates a short narrative — the scent story. It is written fresh for each visitor and changes based on what you were searching for.
The story combines two things: the creator's artistic voice (their philosophy, their signature materials, the aesthetic world they work in) and a bridge to whatever you were looking for when you arrived. If you searched for “something for a rainy city in October”, the story lands the perfume in that feeling. If you came in cold, it opens from the creator's perspective alone.
The story is generated once per visit. Your search phrase is used only in that moment and never recorded.
The creator layer
Scentum treats perfumers as the primary authors of their work. The homepage is organised around 55 individual creators — their aesthetic philosophy, their signature materials, the houses they have worked with.
Creator profiles were written with the help of AI, grounded in their documented history and body of work. The editorial digest that shapes the feed updates daily: a different creator is spotlighted, thematic collections are drawn from the catalogue, and the feed sequence is reordered to surface unexpected connections.
This is not personalisation based on your data. It is editorial curation that shifts with the catalogue itself.
What we collect
We log anonymous events: which product pages are viewed, what searches are run, when the refinement loop is used, and which similar perfumes get clicked. These events are used in aggregate to surface trending perfumes and inform the daily editorial digest.
No personal data is collected. No cookies are set for tracking. No profile is built about you. Your search queries are not linked to your identity. The refinement loop lives only in your browser session.
The real product is the translation layer between data and perception — structured fragrance data, made navigable through conversation.