In a WCDMA system, mobiles select cells based on received common pilot channel (CPICH) signal quality from the different base stations. The cell size therefore depends on radio propagation properties and transmitted CPICH power. For efficient resource utilization, it is important that the traffic is shared fairly evenly over the base stations. This is specifically challenging in irregular networks with inhomogeneous site locations and traffic distribution. This paper discusses methods for automated CPICH power planning, based on differently detailed traffic density information and predicted radio propagation. Simulations using data from a realistically planned network indicate that the tuning results in a more evenly utilized base station power. This in turn improves the overall capacity. It is furthermore fairly robust with respect to traffic density deviations from the true traffic density.