The experimenter effect is one of a number of biases that can creep in to the methodology of a trial that can skew the results before any analysis is even completed on the data. Cues of various types can be given off by a person administering a trial if they are aware of certain conditions or elements of the trial. These cues could be verbal or non-verbal and could be minute, but if picked up by a subject or participant in a trial can affect recorded data and lead to a faulty statistical analysis.
Proper blinding of an experimenter can help to block some of these biases becoming a problem. For example, an independent person in a medical trial will be responsible for giving medication to the people conducting the trial, but there will be no indication whether it is the real treatment or not. Or a person testing responses in a psychological trial will not know what situation is coming up, thereby not giving clues to the subject of the trial.