A new process analysis and quality prediction scheme is presented, based on phase-specific average process trajectory, and is developed for the improvement of quality prediction in multiphase batch processes. After the process trajectory is separated into different phases, based on the different inherent process correlations relevant to quality, quality prediction is performed using average trajectory focusing on critical phases. In this way, it explores the phase-specific cumulation effects of process variables on product quality with a simpler regression model. Then the critical-to-quality phases are identified, and the online quality predicting algorithm is conducted correspondingly during those critical phases. Moreover, to improve the estimation precision of unknown data in each phase, the past batches are made use of, and the missing observations are complemented by searching for the most similar process trajectory to the current one. The applications of the proposed scheme to injection molding show its effectiveness and feasibility. (c) 2008 American Institute of Chemical Engineers.