Evolutionary development allows early and frequent adaptations to new or changed requirements. However, such unanticipated changes may invalidate design documentation and cause structural degradations of the software, which in turn may accelerate changeability decay. Our definition of changeability decay focuses on the increased effort required to implement changes. We have identified three approaches to the assessment of changeability decay: (1) Structure measurement, (2) change complexity measurement and (3) benchmarking. Our research aims to evaluate and compare these approaches in order to develop an empirical assessment framework. In this paper we propose a set of change complexity measures (2) and compare them with structural attribute measures (1) using detailed process and product data collected from a commercial object-oriented development project. The preliminary results indicate that the change complexity measures capture some dimensions of changeability decay not accounted for with structural attribute measures. However, the current findings also suggest that many aspects of changeability decay cannot be accounted for by the indirect measures utilized in approach (1) and (2). As an alternative approach, we therefore propose using benchmarks (3) where change effort can be measured more directly. A research methodology for the development of benchmarks and benchmarking procedures are described.