Hi Daniele, thanks for your response. It’s definitely true that managers are interested in the most relevant variables that better explain a business phenomenon. These variables could become the driver for other activities and could bring some interesting insights for other analysis. Of course there are models that give us their personal representation of feature relevance (GBM, logistic regression and so on), but I always prefer to use an unsupervised approach before training any model. This could remove some bias based on the model’s “way of think” our data.