This research was conducted with the aim of knowing the products that are most in demand by customers, also knowing the products that sell less and knowing the accuracy of testing the performance of sales, stock data. The qualitative method used focuses on interpretation and understanding of interviews or observations to collect sales data, stock of 23 food beverage product items. The results of manual calculation of the k-means algorithm with sales data, stock is to find out the number of products that are in demand by customers with categories: very salable = 3 product items, salable = 5 product items, less salable = 15 product items, then sales data, stock is tested with RapidMiner and Python applications for clustering / grouping what products are in demand by customers the results are the same as the categories: very salable = 3 product items, salable = 5 product items, less salable = 15 product items and looking for the best number of clusters is 3.