Abstract: In a charging system for cellular communications, real-time prices for new connections are offered to the customer. The real-time prices take into account both prevailing and historic traffic patterns, together with the probability that the customer will move from their current cell into a neighboring cell. Cells of the network are monitored in clusters. If a customer is located in one particular cell, the price offered to that customer will take into account the probability of handover to a neighboring cell during the connection, and may also take into account the probability of incoming traffic from neighboring cells during the connection. To enable this predictive pricing, the system stores, and may also update, historic data.