A better way to find your next scent.

How it works

1. 1. Choose your approach, then answer 10 short questions.

   Where you'll wear it, what you already like the smell of, how noticeable it should be, what to avoid and a rough budget band. Everything in plain language.

2. 2. Your preferences shape a shortlist.

   A published weighted base scorer compares your preferences with the catalog. Its Base fit score stays available inside every recommendation. Familiar prioritizes fit; Balanced and Adventurous make more room for discoveries when enough suitable alternatives exist. Owned fragrances remain taste references but leave room for new recommendations.

   See how Base fit is calculated

   • Familiar notes: 30 pts
   • Where you'll wear it: 20 pts
   • Impression: 15 pts
   • Climate: 10 pts
   • Projection: 10 pts
   • Budget: 10 pts
   • Marketing preference: 5 pts

   "Who is this for?", experience level and the optional free-text note are stored as context and do not add points.

3. 3. Notes you avoid stay out.

   Fragrances whose catalog notes match an avoid category are removed before AI compares the shortlist. If that leaves fewer matches, we show fewer. Trying a sample is still the best way to find out how a fragrance feels to you.

4. 4. Your chosen method orders the shortlist.

   Classic uses the fit order. With AI selected, AI looks at notes, character and occasions alongside your answers, then ranks up to 12 shortlisted fragrances. Results waits for a checked response before showing the picks. If AI cannot finish, you get a clearly labeled Classic ranking.

5. 5. Explore your matches.

   See the notes, fit factors and a practical caveat for each pick. Base fit measures the match to your quiz preferences; AI determines the recommendation order when its ranking succeeds. Card explanations come from the catalog's details.

What the AI does

With AI matching selected, ScentScout sends your fragrance shortlist to Lovable AI for a real comparison against your preferences. Classic matching skips the AI request entirely. The response decides the order of your AI-ranked picks. We check that order before showing it or carrying it into your Passport.

Identical fixed answers produce identical base-fit calculations. When qualifying Match Feedback exists, the same answers plus the same signals produce the same locally tuned shortlist of at most 12 eligible records before AI ordering. ScentScout accepts an AI response only when it is an exact ordering of that supplied set, so no new record or hard-excluded record can enter through that step.

The request contains your ten fixed-choice answers and bounded catalog facts: family, notes, tags, seasons, occasions, projection, audience and a coarse price band. The model is not given candidate names, brand names, ScentScout catalog IDs, exact prices, live prices, descriptions, URLs or retailer listings. It is also not given your optional free text, identity, raw account rows or raw Taste Tuning signals and deltas. Local Taste Tuning can still change which eligible candidates reach the shortlist before those bounded facts are sent. Each candidate travels as an opaque position token, which removes direct name and brand cues but is not a guarantee about inference. The ranking call has no external browsing or retrieval tool. Its prompt instructs the model to use only the supplied facts and not prestige, popularity or outside knowledge.

The model returns the recommendation order. Fit scores and card explanations are calculated separately from the catalog record, so you can inspect the same factors across fragrances. An invalid, unavailable or rate-limited AI response leads to Classic ranking, with a short explanation on Results.

When you are signed in and have marked at least one fragrance with Interested, Own it or Not for me, Match Feedback adjusts your Results ranking within an asymmetric bounded range of −10 to +8 points on top of the base fit. Interested is a light boost, Own it is a stronger boost, Not for me reduces similar scents and Tried it is neutral experience context that contributes zero. Only structured catalog attributes are read (family buckets, notes, tags, personality, occasions, seasons and projection): never price, gender, product names, brand names, descriptions or free text. Your quiz "avoid" picks remain absolute. Raw signal rows and numeric deltas are not included in the AI request, although their local effect can change which candidates enter the shortlist.

Retailer links & disclosure

Retailer buttons open a search on the retailer's site: we never point at a specific product listing whose price and availability we can't verify. Retailer links today are neutral, non-affiliate search links. If affiliate links are added later, they will be clearly labeled and will never affect which fragrances appear in your results or where they rank.

Scout a fragrance

/scout is a direct lookup over the same 200-record catalog. Search, catalog facts, Match Feedback, the missing-fragrance request form and the live US price comparison are public and available to every visitor.

A personalized base fit appears only for a signed-in visitor who has a valid saved Passport. Signed-out visitors and signed-in visitors without a saved Passport see no fit score and no quiz or Passport-derived explanation. Your typed search stays in this browser and is never persisted in the URL.

If a fragrance isn't in the catalog, an explicit Request a missing fragrance form lets you submit a short 2 to 160 character note. Nothing is sent until you press the request button; we store only that text plus a timestamp and never ask for an email.

An optional Live US price scout panel queries Google Shopping for full-bottle listings in the US. The browser sends only the catalog fragrance ID to a same-origin ScentScout endpoint; the API key stays server-side. Samples, decants, gift sets, testers, oils, body sprays, aftershaves and dupes are filtered out before ranking. Offers are grouped by exact bottle size and concentration; only the eligible minimum with a known delivered total is badged as the lowest for that variant. Shipping is treated as unknown unless the retailer explicitly reports it; tax is excluded; seller authenticity is not verified. When the API is unconfigured or errors, the panel falls back to a Google Shopping compare-all link and neutral retailer searches.

Who runs it

A project by Dibs. If something's wrong or feels off, that's on me. I'd rather know.