Controls are always the toughest issue in experimental research.

http://norvig.com/experiment-design.html