Recommendations are crucial for the success of large websites. While there are many ways to determine recommendations, the relative quality of these recommenders depends on many factors and is largely unknown. We propose a new classification of recommenders and comparatively evaluate their relative quality for a sample website. The evaluation is performed with AWESOME (Adaptive website recommendations), a new data warehouse-based recommendation system capturing and evaluating user feedback on presented recommendations. Moreover, we show how AWESOME performs an automatic and adaptive closed-loop website optimization by dynamically selecting the most promising recommenders based on continuously measured recommendation feedback. We propose and evaluate several alternatives for dynamic recommender selection including a power-ful machine learning approach.