How our science works

Where the notes come from

• Starter library. Each of the 190 starter fragrances was checked against the brand’s own page or a major retailer’s product page, and every fragrance page links to that source.
• Automatic lookups. Fragrances not in the library can be looked up automatically. Those entries are labeled as such and may contain mistakes until someone corrects them.
• The community. Anyone signed in can correct notes and attach a source link. Every change is logged with the before and after.
• We don’t copy from fragrance databases such as Fragrantica, apart from one starter entry whose brand page lists no notes.

What we estimate, and from what

• Scent families. Each listed note is sorted into one of 12 families (citrus, fresh, aromatic, floral, fruity, gourmand, amber, spicy, woody, smoky, oud, musk). Base notes count most because they last longest; top notes count least.
• Weight. Each family has a heaviness, from bright citrus to dense oud and resin. A bottle’s weight is the average of its notes, scored 1–10.
• Longevity and sillage. Starting estimates come from the concentration (Eau de Toilette, Eau de Parfum, Extrait and so on) adjusted for weight. Denser bases tend to last longer and project further.
• Heat and cold. Heat speeds evaporation, so scents project harder early and fade sooner, and sweet, dense compositions can turn cloying. Lighter scents start with a higher heat estimate.
• Layering. Two bottles pair well when they share a note (the “bridge”), when one is clearly heavier than the other, and when they come from different families. The heavier one goes on first. Two dense scents together are flagged to avoid.

What we measure

• Wear logs. After wearing a bottle, people record how many hours it lasted, its sillage (intimate, moderate, strong or room-filling), how it held up, and the temperature. The weather check fills the temperature in.
• Combo ratings. People who try a combo rate it 1–5 and say whether they’d wear it again.

How the two are combined

• Each estimate counts as if 3 people had reported it. As real reports come in, they outweigh the estimate. After 3 wearers a figure is labelled measured; after 30 the estimate barely matters.
• Each person counts once per bottle. Logging the same scent fifty times doesn’t move the average more than logging it once.
• Heat scores only use wears at 85°F or warmer; cold scores only use wears at 55°F or cooler.
• Well-rated combos rise in everyone’s rankings and poorly rated ones sink, even when the rules predicted otherwise.

Known limits

• Note lists are marketing, not formulas. Brands keep their real recipes secret, and the molecules that drive performance (such as Ambroxan, Iso E Super and synthetic musks) often aren’t listed.
• Mixtures aren’t additive. Smell research finds one scent can fully mask another, and blends of more than about three odors tend to be perceived as one new whole. That’s why layering predictions get tested by wearers rather than trusted.
• Skin and batches vary. Skin type, hydration and batch differences change performance. Averages across many people smooth this out; a single wear doesn’t.

Sources: Rodriguez-Raecke et al., “A Masked Aversive Odor Cannot Be Discriminated From the Masking Odor…”, Frontiers in Neuroscience, 2019. Spotted something we should change? Correct the notes on any fragrance page, or log your wears. Back home