In our prior work, we presented a novel approach to the evaluation of quality in use of corporate web sites based on an original quality model (QM-U) and a related methodology (EQ-EVAL). This article focuses on two research questions. The first one aims at investigating whether expected quality obtained through the application of EQ-EVAL methodology by employing a small panel of evaluators is a good approximation of actual quality obtained through experimentation with real users. To answer this research question, a comparative study has been carried out involving 5 evaluators and 50 real users. The second research question aims at demonstrating that the adoption of the EQ-EVAL methodology can provide useful information for web site improvement. Three original indicators, namely coherence, coverage and ranking have been defined to answer this question, and an additional study comparing the assessments of two panels of 5 and 10 evaluators, respectively, has been carried out. The results obtained in both studies are largely positive and provide a rational support for the adoption of the EQ-EVAL methodology.