Eurostat implements two different kinds of user satisfaction surveys over the Internet – general user satisfaction surveys and themed user satisfaction surveys. Whereas the general survey takes place periodically (every year or every two years), the themed satisfaction surveys are conducted at irregular intervals within the framework of the evaluation of statistical production processes. This paper focuses on and develops the pros and cons of both in order to learn more about users, their needs and their satisfaction with the services provided by a statistical institute. The paper is based on the experience with three Eurostat's general user satisfaction surveys (2007, 2009 and 2011) and on the themed surveys conducted in about 20 statistical domains, mostly linked to rolling reviews. The main shortcomings of the general surveys, conducted using standard questionnaires answered via Internet, are the representativeness of the overall population and the difficulty in making comparison of results over time, caused by organisational changes, modifications in priorities, product evolutions. The principal problem with themed surveys is how to cope with limited numbers of answers. Although themed surveys use tailor made questionnaires, there are common elements that allow comparing some outcomes and establishing common patterns for these kinds of surveys. Finally a number of recommendations will be provided on how to improve different quality aspects of these surveys.