The paper addresses the e -commerce system (interpr eted here as information system - IS) understandability problem from the maintenance and evolution perspective. We propose a methodology that includes: (1) identification of the 3 representative systems (through clustering of a set of previous ly developed IS on the basis of user-oriented and temporal criteria ); (2) extraction of feature models from representative systems for evaluation using similarity (measured by absolute difference val ue - ADV) and complexity metrics leading to the definition of the reference IS - RIS; (3) construction of the business logic feature model, represented as meta -graph, to reason about quality of the systems and to understand evolution trends; (4) feature mo del refactoring / modularization of the RIS ; (5) Code-level modularization of the RIS. The methodology is supported by experiments to evaluate the analysed problems. The basic result s are: (1) feature models of the representative systems; (2) metrics to ev aluate complexity and similarit y of the represented systems; (3) the re -engineered RIS to support better maintainability and design procedures.