Le Labo, an artisanal perfume brand owned by the beauty giant Estee Lauder, brings a fresh approach to the perfume industry. They celebrate and promote the craftsmanship of scents, which has been arguably spoiled by the big luxury brands (Why does everyone smell like a Sephora store?). Le Labo has a qualitative and out-of-the-box questionnare to (Called “Proust Questionnare”, Exhibit 1) identify each individual’s taste in scents, and then hand-create and deliver the bottle to customer. Hence, their ultimate value proposition is to match a personality with a scent, and then deliver a “one of a kind” bottle created by hand for you . While Le Labo goes against the grain in terms of technology and innovation, could there be an opportunity to leverage technology to boost personalization in the fragrance industry?
Exhibit 1 – Proust Questionnare (selected questions)
Personalization has become a mega trend in many industries, but most importantly for the beauty industry . How human experience different scents is one of the most complicated notions to conceptualize, and depends on many complex factors such as our genetics, experiences and culture . This is why machine learning could add tremendous value.
IBM has already invested in extended research in the perfume industry, specifically with their AI Philyra. Philyra replicates the fragrance making process, just like a human apprentice, and crafts fragrances for different segments of customers . This is an extremely important step towards analyzing and understanding fragrance formulas and demographic preferences. However, Le Labo would not care about mass segments, since they are trying to deliver a unique and hand-crafted product. The question then is how can AI resolve the link between numerous personal traits and taste in scent?
Le Labo (and Estee Lauder) tackles this opportunity in personalized scents in two ways. Firstly, the Proust Questionnare helps identify unique personal traits that are impossible to codify and analyze as structured data (such as your weirdest extravagance). The answers to this questionnare is assessed by a team and then a selection from the existing fragrance portfolio is recommended to the customer. This process is currently mostly manual or rule-based. Secondly, Estee Lauder collaborates with IBM in the Philyra project by providing its fragrance database.
Can Estee Lauder take a bold step and create their own AI for the ultimate personalized fragrance creation? The straightforward answer might be that they cannot because they do not have the technical capabilities; however, it is crucial to note that they are sitting on a tremendously valuable data, retrieved from their large brand portfolio  . Because of their “own-store” retail strategy for their most brands, they know which shade of lipstick that I purchase from MAC, how often and where I shop, which skin problems I have, which perfumes that I previously purchased and repurchased. These might not seem like a direct relationship points for an individual’s taste in fragrance, and this is exactly why AI could bring tremendous insight. Combining all previous product purchases together with product attributes (e.g., style, functionality, scent, color) and shopping behaviors (e.g., location, time, frequency, social vs. individual) will create an input data with 10,000+ columns, which could be or not at all related to a person’s taste in scent. But that is for the AI to decide. Then the next step is crafting the output. Let’s say each fragrance have about 100 attributes (e.g., top notes, base notes, longevity), so AI has to identify the optimal combination of these attributes.
In the short and medium term, considering the infinite number of combinations and potential complications in operations, Le Labo could try to cluster their existing portfolio of scents to match these personal attributes that we previously mentioned. Following this, they could send free samples to Estee Lauder’s loyal customers and ask them to review them as a feedback mechanism back to the AI. The ultimate personalization of the fragrances will be a longer term application, combined with advancements in neuroscience, biotechnology and chemistry.
Perfume is an individual’s most personal choice, and many dare to call this industry an art rather than science. Hence, when we introduce machine learning, would the true purist customers, who want an authentic and hand-crafted experience, embrace this new trend? If not, should this technology rather be non-transparent to the consumers? Or for the data enthusiasts like myself, would these consumers be willing to share more personal data such as their social media profiles or book/music playlists to help this AI achieve the maximum personalization? And rather a bigger question for the perfume industry, if the fragrance search is over after one try with AI, would not this significantly shrink the beauty industry?
 “Le Labo”. 2018. Elcompanies.Com. https://www.elcompanies.com/our-brands/le-labo.
 Council, Young. 2018. “How Personalized Goods Are Shaping The Economy”. Forbes. https://www.forbes.com/sites/theyec/2017/05/05/how-personalized-goods-are-shaping-the-economy/#1457e1b23a1c.
 “Smells Like Machine Learning Progress – Towards Data Science”. 2018. Towards Data Science. https://towardsdatascience.com/smells-like-machine-learning-progress-611a2851acec.
 “Forget About Chanel No. 5. IBM Is Now Making Perfume Using AI.”. 2018. Vox. https://www.vox.com/the-goods/2018/10/24/18019918/ibm-artificial-intelligence-perfume-symrise-philyra.
”Our Brands”. 2018. Elcompanies.Com. https://www.elcompanies.com/our-brands.