This study aims to identify the root causes of customer churn in the telecommunications industry and to predict churn using network-related indicators. Unlike many previous studies that rely on billing or campaign-related variables, this work focuses solely on network performance metrics (KPIs) that reflect real customer experience — including streaming quality, end-to-end delay, VoLTE QoE, and overall network quality indicators. As the dataset does not contain any economic or tariff-related information, the results specifically reflect the impact of non-economical churn drivers. A 30-day average of daily KPIs was calculated for each customer across 2G, 3G, and 4G technologies, and a representative sample was selected for clustering. Bisecting K-Means algorithms were applied to segment customers. The segment identified with poor network performance showed significantly higher churn compared to other groups, revealing a strong link between degraded network experience and churn behavior. Subsequently, ANOVA was used to identify the most influential features on churn, and several machine learning models were trained using only the statistically significant variables. The models' precision and recall scores were compared to determine the best-performing model. As the initial results were not satisfactory, a more focused case study was conducted using 30-Day Daily Time Series per Customer data for the Kadıkoy, Sisli, Uskudar, Besiktas districts of Istanbul. Churn prediction was then performed on a balanced dataset, created by sampling an equal number of non-churners to match the churners in the same period, achieving an F1-score of 62 %.