Online platforms allow customers leave their opinions after using products or services. Data collected from customers is enormous. How to organize and transform the data to useful information is one of the most important work to improve future business strategies. This study focuses on customer analytics to define the potential certain groups of customers so restaurants can concentrate their efforts and to build a database system that includes a satisfaction rating scale in order to benefit in the customer's experience. This study applies a method combining machine learning (ML) techniques and multi-criteria decision making (MCDM), with the case study of customers on Foody Vietnam. Using the criteria rating data, the revised Fuzzy C-Means (RFCM) clustering technique builds descriptive modeling to discover customer segmentation while Random Forest (RF) technique builds predictive modeling to show satisfaction classification. Method Based on the Removal Effects of Criteria (MEREC) ranks the criteria that customers of each generated segment rate on the Foody platform, which are Location, Space, Quality, Service, and Price. The results give three segments of customers with different criteria ranking and a customer preference prediction. The study provides a better insight into customer understanding and therefore, helps restaurant owners in terms of improving overall customer experiences.