<![CDATA[Fluctuations in the demand for medical devices can trigger the risk of stock shortages (stockouts) and overstock conditions, which may affect operational costs and the quality of distribution services. This study aims to classify medical device products at CV Patriot Kencana Medika Kudus based on demand patterns and purchasing characteristics, and to map the clustering results as an initial basis for developing purchasing schedules. The data used consist of internal purchasing transaction histories from the 2023–2025 period with four main features: Quantity, Price_Per_Unit, Lead_Time_Days, and Total_Purchase_Value. The methods applied include exploratory data analysis, feature construction and normalization, determination of the optimal number of clusters using the Elbow Method and Silhouette Score, K-Means modeling, and evaluation using the Silhouette Score and Davies–Bouldin Index (DBI). The results indicate that the use of three clusters provides the most reasonable compromise between the inertia reduction pattern, Silhouette value, and managerial interpretability. A Silhouette Score of 0.2563 and a DBI value of 1.349 suggest that the quality of cluster separation remains at a low to moderate level, meaning that the resulting clusters are more appropriately interpreted as an initial segmentation rather than a fully distinct classification. The three clusters formed were interpreted as general products, premium products, and strategic products. The numerical characteristics of each cluster were then used to calculate simple indicators, namely the reorder point (ROP) and economic order quantity (EOQ), as baseline purchasing recommendations. The main contribution of this study lies in integrating clustering results with operational inventory policy parameters, although the findings still need to be interpreted cautiously because they have not yet been compared with other algorithms, their stability has not been tested, and the EOQ model applied remains simplified.]]>
