This work describes a method for the "analysis and prediction of performance for evolving architectures " (APPEAR). The method aims at performance estimation of adapted parts of software product families during the architecting phase. It combines both structural and statistical techniques in a flexible way, that is, it allows choosing which part of the component is structurally described, modeled and simulated, and which part is statistically evaluated. The method was exemplified by case studies in the consumer electronics and the medical imaging system domains. The results of the initial validation of the method are encouraging.