Customer churn is a major challenge in the telecommunications industry, where keeping existing users is more cost-effective than gaining new ones. This study introduces a novel, time-aware approach to churn prediction using eight years of data from over 8,000 customers of a Polish telecom provider. Unlike most studies that rely on fixed or aggregated data, we apply the Cox proportional hazards model with time-dependent variables to track how churn risk changes over time. A key innovation is the combination of static demographic data with detailed, time-varying service usage—an approach rarely used in churn modeling. This allows for a deeper and more accurate understanding of customer behavior. Our results show that increased use of calls, minutes, and MMS lowers churn risk, while more SMS usage slightly raises it. Customers from more populated areas also tend to stay longer. This dynamic and integrated method offers a new framework for churn analysis and supports more effective, data-driven decisions in customer retention and business intelligence systems.