The importance of companies' website as instrument for relationship marketing activities is well-known both in the academia and in the industry. In the last decades, there has been great interest in studying how technology can be used to influence people’s attitudes and motivate behavior change. With this, web personalization has had increasing research and practitioner interest. However, the evaluation of user interaction with companies' websites and personalization effects remains an elusive goal for organizations. Online controlled experiments (A/B tests) are one of the most commonly known and used techniques for this online evaluation. And, while there is clearly value in evaluating personalized features by means of online controlled experiments, there are some pitfalls to bear in mind while testing. In this paper we present five experimentation pitfalls, firstly identified in an automotive company’s website and found to be present in other sectors, that are particularly important or likely to appear when evaluating personalization features. In order to obtain the listed pitfalls, different methods have been used, including literature review, direct, and indirect observation within organizations of the automotive sector and a set of interviews to organizations form other sectors. Finally, the list of five resulting pitfalls is presented and some suggestions are made on how to avoid or mitigate each of them.