InstaCart Analysis: Use PCA with K-Means to Segment Grocery Customers

Researching customer classification can effectively help businesses predict future buying trends and help customers have a better purchase experience. The study can be applied to major retail enterprises to help them improve the payment conversion rate and order rate at the same cost. This paper uses InstaCart as the subject of the study and analyses its customer orders for three years. The classification results of the study to describe each clusters’ characteristics and help enterprises maintain the best level of inventory supply. The study is based on the Gold Award python notebook of participant Andrea Sindico. Principal Component Analysis is a machine learning approach in various applications. This paper aims to use PCA to find new dimensions and to cluster the customers by their purchase behaviour. After analysis, this study only keep the top 6 key component and chooses two best-selling of the six aisles (PC1 and PC4). The study resulting in four different clusters for customer segmentation, and different clusters have their unique characteristics for customers’ future orders.