The hackathon of the chair « Next Gen RetAIl » comes to an end
The annual hackathon organised by the chair of the École polytechnique « Next Gen RetAIl » has, for this second edition, gathered 132 students organised in 26 teams. This challenge of data science applied to retail lasted from the 3rd of February to the 8th of April. Carrefour, sponsor of the chair, lent some of its data to create three challenges. This first one was about how to optimise the “DNA product” (price + product) on a store scale. The second one was about designing a local marketing Mix for the betterment of local marketing strategies. The third one was about optimising the frequence and packaging of the in-store supplies.
To deal with these challenges, participating students had access to thirty data records on products, pruchases by client, marketing investments and sales from five stores in the Côte d’Or department (anonymised data).
The event was separated in two phases : for two weeks, the teams analysed the data and proposed a solution idea. This preliminary work allowed the selection of the twelve teams with the best project, who were then able to develop their project for the rest of the event. Weeks of hard work were shouldered by a Carrefour DataLab mentor and ended with a pitch session, held by a jury of directors of the brand.
The winning team was comprised of Telecom Paris of the Institut Polytechnique de Paris and is named VapnickLovers in reference to Vladimir Vapnick, mathematician and informatician, famous for his contribution on the statistical learning theory. To answer the second challenge, they chose to focus on the consumption of organic products, which sales revenue shoud reach five billion euros in 2022. To make their results more realistic, they enriched Carrefour’s data with other elements with an impact on consumption, such as the “Gilets Jaunes” movement, or the presence of the nutri-score.
MixLocal.ia was the result of their work. It is a machine learning software combining two algorithms designed to optimise the two criteria. The first algorithm is very effective to treat data with lots of categories, such as in this case with all the different products and their shelves. The second one highlights the highest impact variables, like the presence of one product or the use of specific marketing media in this case. Thus, MixLocal.ia should be able to measure the impact on sales on a store-scale of the presence or absence of a given product or the use of a specific marketing media. It has the objective of being the spearhead of food transition.
About the chair « Next Gen RetAIl »
Held since 2019 by Eric Moulines, member of the Academy of Science and professor at the École polytechnique de l’Institut Polytechnique de Paris, the chair « Next-Gen RetAIl » encourages research and teaching in Artifical Intelligence and Data Science. It develops algorithms that can be used in retail. Supported by Carrefour and interacting with its Digital HUB, the chair fits in the scope of these thematics through its research and teaching oriented towards conversational AI and product recommendation systems.