In order to improve the innovation process, concurrent engineering may be associated with knowledge capitalization methods. The objective is to reuse knowledge having emerged during previous innovative projects. This article brings a methodological contribution to project feedback within new product development tasks. First, the interests of project feedback activities are underlined. More particularly, the impact of project feedback on knowledge management cycles is discussed. Then, a methodology based on a specific information extraction approach from past projects is detailed. The TRIZ theory constitutes the theoretical background of this research. An experiment in the field of machinery development is presented. To conclude, this article discusses the assets and limits of the approach and gives some practical advice concerning the use of memorized data within innovative projects.