<![CDATA[Abstract. Increasing business competition requires companies to understand customer behavior in order to maintain customer loyalty and reduce the risk of customer churn. This study aims to perform customer segmentation and identify customers with churn potential using the K-Means Clustering method based on Recency, Frequency, and Monetary (RFM) values. Transaction data were obtained from PD XYZ, a store operating on an e-commerce platform in Indonesia. The data were processed through preprocessing, transformation, and scaling using RobustScaler to standardize the data and reduce the influence of outliers. The optimal number of clusters was determined using the Elbow method and evaluated with the Silhouette Coefficient and Davies–Bouldin Index. The results indicate that k = 3 is the optimal number of clusters, with a Silhouette Coefficient of 0.771495 and a Davies–Bouldin Index of 0.547348. The segmentation produced three customer groups: non-loyal, moderate, and loyal customers. A total of 3,840 customers were identified as having churn potential, while 236 customers were classified as non-churn. The findings demonstrate that the RFM-based clustering approach is effective for understanding customer characteristics and supporting the development of more targeted customer retention and marketing strategies. Abstrak. Peningkatan persaingan usaha mendorong pelaku usaha untuk memahami perilaku pelanggan guna mempertahankan loyalitas dan mengurangi risiko kehilangan pelanggan. Artikel ini bertujuan melakukan segmentasi pelanggan serta mengidentifikasi pelanggan yang berpotensi churn menggunakan metode K-Means Clustering berbasis Recency, Frequency, dan Monetary (RFM). Data transaksi yang digunakan berasal dari PD XYZ, sebuah toko yang beroperasi pada platform e-commerce di Indonesia. Data diproses melalui tahap pra-pemrosesan, transformasi, dan scaling menggunakan RobustScaler untuk menyesuaikan skala data serta mengurangi pengaruh outlier. Jumlah klaster ditentukan menggunakan metode Elbow dan dievaluasi dengan Silhouette Coefficient serta Davies-Bouldin Index. Hasil menunjukkan bahwa k = 3 merupakan jumlah klaster terbaik dengan nilai Silhouette Coefficient sebesar 0,771495 dan Davies-Bouldin Index sebesar 0,547348. Segmentasi menghasilkan tiga kelompok pelanggan, yaitu pelanggan tidak loyal, pelanggan moderate, dan pelanggan loyal. Sebanyak 3.840 pelanggan teridentifikasi berpotensi churn, sedangkan 236 pelanggan tidak churn. Hasil penelitian menunjukkan bahwa pendekatan RFM berbasis klaster efektif dalam memahami karakteristik pelanggan dan mendukung penyusunan strategi retensi serta pemasaran yang lebih tepat sasaran.]]>
