the answer is : having more than one independent variable
If you have more than one independent variable, you won't be able to find out which variable that actually produce the end Result which make the experiment lose its "controlled" element and make the result really unreliable