Abstract-Knowing how a software project is likely to evolve is an essential problem for any software project manager. Finding efﬁcient predictors for performance indicators (e.g., bugrates) has been the focus of many studies. Previous studies found that process metrics make likely candidates for this predictor role, for bug data in particular. We propose a methodology for in-depth analysis of process metrics to ﬁnd out how they relate to changes. We use a lexical approach to classify changes into perfective, adaptive and corrective changes. The analysis consists of examining a set of hypotheses on the nature of the relationship of certain process metrics and the change categories, e.g., perfective changes and their impact on the consecutive bug appearance rate of a module. As this work is in progress, we present a pilot study on a module of the Ant Project to showcase and discuss our technique and point out early trends.