Product derivation approaches automate the customization process of software product lines. Over the last years, many tools have been proposed aiming at synthesize and generate products from a set of reusable assets. These tools adopt different techniques and strategies to implement and automate the product derivation activities. In this paper, we analyzed six modern product derivation tools (Captor, CIDE, GenArch, MSVCM, pure::variants, XVCL) in the context of evolution scenarios of a software product line. Our study has adopted several metrics to analyze the modularity, complexity and stability of product derivation artifacts related to configuration knowledge along different releases of a mobile product line. The preliminary results of our study have shown that approaches with a dedicated model or file to represent the CK specification can bring several benefits to the modularization and stability of a software product line.