Questionnaires for the WWW are easy to plan and can be distributed over a large area. Carrying out surveys being that simple, one can easily forget that questionnaires are an artificial form of communication and crafting them is an art to be learned and mastered. The quality of online surveys has improved increasingly, yet mistakes that cause poor quality of data are still frequent. When analyzing mistakes, one must differentiate between mistakes made by beginners with little or no experience of survey techniques and mistakes caused by the translation of normal techniques to the new medium. Bearing in mind the fact that online surveys have only been possible since the middle of the nineties, it is clear why as yet no systematic knowledge of the possibilities of conducting WWW surveys exists. The University of Cologne has developed a virtual pretest studio to test and optimize WWW surveys before they are introduced into the field. On the basis of the experiences made there, this contribution will outline some of the more common mistakes and develop guidelines for valid online surveys. Furthermore, the online pretest studio will be introduced as a quality control device.