<![CDATA[Background: In online marketplaces where thousands of products compete with similar prices and ratings, sellers struggle to decide which products deserve promotion, retention, or removal. As competition in Indonesian e-commerce intensifies, data-driven approaches are needed to manage product performance. Objective: This study aims to formulate a product management strategy using a two-stage data-driven approach. It focuses on performance-based product segmentation and classification and identifies the key variables that show the strongest relationships with performance segments and sales outcomes. Method: This study uses a census-type sample of 2,547 household products in the “Rumah Tangga” category on Tokopedia, collected via web scraping in April 2025. K-Means clustering segments products based on product rating, total product ratings, store rating, and total store ratings; Random Forest classifies products into the identified segments; and correlation analysis examines relationships between attributes, performance segments, and sales outcomes. Results: The segmentation analysis produced three product performance segments: low, medium, and high. The Random Forest classifier categorized products into these segments with 99.4% accuracy. Correlation analysis indicates that product and store ratings play a central role in differentiating performance segments, while the number of product ratings is more closely associated with sales outcomes. Conclusion: The findings support strategies such as targeted promotions for high-performing products and inventory adjustments by segment, while strengthening customer rating engagement and store reputation to improve product performance signals. This study extends the literature on data-driven product management by demonstrating how a combined segmentation-classification approach can operationalise performance-based portfolio decisions in an emerging market context. Keywords: data-driven strategy; e-commerce; machine learning; product performance; segmentation]]>
