Promotion Strategy Optimization with Sales Data Analysis: Comparison of K-Means, K-Medoids, and Gaussian Mixture Models for Customer Segmentation

In a data-driven business landscape, understanding customer behavior is vital for optimizing marketing strategies. Generalized promotions, conducted without insights into transactional patterns, often lead to wasted resources and low conversion rates. This study aims to segment customers based on their transaction behavior and identify the optimal time for targeted promotions. Three clustering algorithms K-Means, K-Medoids, and Gaussian Mixture Model (GMM) are compared due to their distinct approaches to defining cluster centers. The research utilizes transactional data from a company, enriched with detailed attributes such as product type, brand, quantity, and price. Data preprocessing was performed using Excel, while analysis and modeling were conducted in Python with Jupyter Notebook. Clustering quality was evaluated using the Davies-Bouldin Score. Results reveal that K-Means achieved the best performance with a score of 0.52. Cluster 2 emerged as the most responsive segment, with 19:00 identified as the optimal promotion time. This study underscores the impact of tailored marketing through customer segmentation and time-based targeting, offering actionable insights that enhance promotional effectiveness, reduce marketing inefficiencies, and drive sustainable sales growth.