How do problem domains impact software features? We mine software code bases to relate problem domains (characterized by imports) to code features such as complexity, size, or quality. The resulting predictors take the specific imports of a component and predict its size, complexity, and quality metrics. In an experiment involving 89 plug-ins of the ECLIPSE project, we found good prediction accuracy for most metrics. Since the predictors rely only on import relationships, and since these are available at design time, our approach allows for early estimation of crucial software metrics.