Association Rule Mining (ARM) is a Machine Learning technique widely used to enhance the predictability of bundling goods and services purchased together. In the competitive world of coffee shops, providing enticing and customized product combinations is crucial for enhancing customer experience and building loyalty. We utilized a newly generated transactional database from a coffee shop franchise in Mexico to identify unique itemsets using a standard Apriori principle. Furthermore, our main contribution is the application of a Two-Phase High Utility Itemset algorithm to predict tailored baskets based on both frequency and monetary revenue simultaneously. Whether suggesting a well-rounded breakfast or a light lunch, smart recommendations can encourage firms to explore new product options and increase their average revenue.