Advancements In Perfumery (AI predicting how strong a perfume material smells)

🔥

Another new area in perfume science is AI trying to predict how strong a perfume material will smell just by looking at its molecular structure.

This study was not conducted by one of the big perfume houses, such as Givaudan or Firmenich. It was an academic machine-learning study by Peter Fichtelmann and Julia Westermayr. They used public odor information from sources like The Good Scents Company and PubChem to build a larger data set of more than 2,300 perfume-related molecules, then trained AI models to predict whether a material would smell odorless, weak, medium, or strong.

The researchers found that things like molecular size, weight, shape, polarity, rings, and branching can help predict whether a material may come across as weak, moderate, strong, or almost odorless. The study also showed that odor strength is not always easy to separate into perfect categories, because some weak and strong materials can overlap chemically. That shows how complicated smell really is.

To me, this is important because it proves performance is not just about “more oil.” A fragrance can have a high oil concentration and still not project as people expect. On the other hand, a tiny amount of the right material can fill a room. That is because the molecule itself matters.

A material has to evaporate off the skin, travel through the air, survive long enough to be noticed, and reach the smell receptors in your nose. If a molecule is too heavy, too sticky, too soft, or does not diffuse well, it may sit close to the skin even if the fragrance is expensive or highly concentrated. If a molecule has the right balance of volatility and diffusion, it can feel loud, airy, radiant, or long-lasting.

This also helps explain why some perfumes smell powerful in the air but not heavy up close, while others smell rich on skin but do not project much. Projection, longevity, sillage, and strength are related, but they are not all the same thing. A fragrance can last a long time and still be quiet. Another fragrance can project hard for two hours and then fade faster.

That is why I do not believe performance should only be judged by oil percentage or whether something is EDP, extrait, or parfum. Concentration matters, but the materials, structure, balance, and evaporation curve matter just as much. Perfumers are not just dumping in more aroma chemicals. They are building a formula where each material has a job: lift, diffusion, body, smoothness, trail, or staying power.

That makes this study interesting because it shows AI is starting to look at perfume materials from the chemistry side, not just the marketing side or the notes pyramid side. It is trying to understand why one molecule can be powerful in tiny amounts, while another molecule may need more material and still sit closer to the skin.

This kind of AI research could eventually help perfumers screen materials faster, understand which molecules may have more impact, and design fragrances with a better balance between strength and beauty. But it also shows the limits of technology. AI can help predict patterns, but smell is still personal. Skin chemistry, temperature, humidity, dosage, blending, and the person smelling it all change the final experience.

So when someone says, “This fragrance is weak,” it may not always mean poor quality or low concentration. Sometimes the materials are designed to sit closer. Sometimes the fragrance is built for elegance instead of beast mode. And sometimes the chemistry just does not push loudly off that person’s skin.

That is what makes perfume so interesting to me. It is not just art, and it is not just science. It is both. The beauty is in how the chemistry, the perfumer’s creativity, and the wearer’s skin all come together.

2 comments