Which experiment method is important for controlling bias?

Randomizing the unwanted variables is one method of building a stronger causal argument. Controlling or a strong attempt at controlling the unwanted variables would be recommended. One variable, and only one should remain independent; this would ensure the dependent variable could be assessed in the proper light. Eliminating unwanted confounding variables my be necessary for a stronger causal argument; the confounding variables distort the conclusion in the causal argument. Eliminating unwanted variables could mean categorising data; it could mean separating data; it could mean some guess work, such as adding/subtracting figures like a statistician.