With decades of development, data warehouses architectures (DWHAs) have been extended to a variety of derivatives. They have been built in multiple environments for achieving different organisations’ requirements. Data warehouse projects may undergo disparate stages during their life cycle (e.g. design, development, maintenance, melioration, migration and discarding). In each stage, some issues may arise to obstruct operations or new requirements may need to be achieved in order to address new challenges and competition. It is therefore necessary to evaluate DWHAs to obtain insights into these derivatives. However, limited research studies have been conducted on how to explicitly and efficiently evaluate DWHAs. They normally require experts for support during evaluation, which makes the processes involved time-consuming and costly. A diversity of features are retrieved and measured in previous research studies when evaluating DWHAs. As a consequence, the evaluation results may be inconsistent and inadequate, due to much manual intervention under different scopes of DWHAs and features. This research proposes an explicit and efficient method for DWHA evaluation. This method is aimed at reducing the ambiguity of results and the interpretation from experts. It is based on two methods also proposed in this research which systematically classify representative DWHAs and identify reliable features. These DWHAs and features are frequently and widely concerned in the domain of DWHA research, which are further interpreted and organised to scientifically evaluate the modelled DWHAs. Furthermore, the evaluation method is implemented with these interpreted DWHAs and features. Finally, this method is evaluated in multiple cases and empirical research through industry and experiments. Consequently, this research outputs a methodology with three methods and makes contributions to facilitate academia and industry on feature identification, DWHA modelling and evaluation.