Cdiscount is the French leader on e-commerce, with millions of visitors and millions of products sold. But that is not all: Cdiscount is also a company where dozens of Data Scientists are daily challenged with a massive amount of data coming from visitors, customers, its product catalogue and internal search engine. Their goal: deliver performing and innovating algorithms in a short delay. The Cdiscount’s Data Science teams use Python for both prototyping and production, trying to shorten the path between a research idea and its industrialization.
In this talk, we are going to present some examples of Python applications in Cdiscount and we are going to focus on one in particular: the keyword clusterisation in Google Ads campaigns using the Word-2-Vec algorithm of the gensim package. The solution of this NLP problem allows to use the signal of high-traffic keywords to predict the performance of lower-traffic ones and improve the bidding.