The science and art of creating fragrances, aromachology, is an ancient and prized one. The world’s most renowned makers of perfumes, aftershaves and cosmetics pay handsomely for the best perfume chemists. These ‘noses’, as they are referred to in the industry, spend years learning their craft, studying under ‘masters’. The establishment of historically elite ‘fragrance agencies’ can almost name their price.
The skill of creating fragrances and aromas is so valued for good reason. Of all of our senses smell is most closely connected to our emotional memory. A smell can trigger buried memories, stir up emotions and even change our mood. Numerous studies have demonstrated our attitude towards things, including buying behaviour, is most closely connected to emotion. We might like to think that most of the decisions we take are based on reason. But the evidence suggests that in the majority of cases it is in fact our emotions lead us to a conclusion that we then find rational support for rather than the other way around.
The emotional influence of our sense of scent is the basis of the perfume industry and why it has been big business for so many years right back to antiquity. But the industry may be about to undergo the most significant disruption in its history. IBM has developed an AI algorithm that it believes will allow it to create the perfect fragrances. Will it mean that the ‘noses’ of the traditional perfume industry will soon be looking for new employment opportunities? It is, of course, too early to say but the development is certainly an intriguing one.
The AI development, and IBM’s involvement, has actually been instigated from inside the perfume industry. Symrise is a German fragrance house whose stable of clients include names from Estée Lauder to Avon. The company approached IBM with the suggestion of attempting to apply the IT giant’s AI technology to perfume fragrances. The algorithm developed has been named Philyra and develops fragrances for specific market segments. Two unnamed fragrances have already been sold to Brazilian beauty products chain O Boticário and are already on its shelves.
Philyra work is based on a dataset of around 1.7 million fragrance formulas developed by Symrise as well as sales data of the different perfumes they have been included in either as a mix or independently. The latter includes the geography, demographic and socio-economic information of buyers as well as purely sales numbers. Based on the information Philyra comes up with new fragrance recipes based on a particular demographic, such as Brazilian millennials.
Market research conducted around the two fragrances acquired and marketed by O Boticário suggest the AI approach shows promise. Focus groups loved them and they were favoured ahead of other perfumes they were asked to judge. The they are also said to have become top sellers after hitting the shelves.
However, traditional perfume ‘noses’ are not on the verge of being replaced by AI just yet. While IBM’s algorithm came up with the main formulas of the two perfumes selling now, those were then ‘tweaked’ by a master perfumer. At least initially the main role of Philyra will be helping Symrise improve the speed and efficiency at which it comes up with new scents.
AI is considered to have broad potential application across the fashion and beauty industry. Fashion labels and retailers have started experimenting with its potential around predicting what styles will sell over the next season and there are also skincare start-ups developing AI algorithms to come up with personalised skin creams that are specifically to individuals.