A confounding variable is associated with both the exposure and the outcome of interest.

For instance, alcoholics are more likely to get lung cancer. You might conclude that alcohol causes lung cancer. However, alcoholics are often heavy smokers and smoking causes lung cancer. Here, smoking is associated with both being an alcoholic and developing lung cancer. Smoking is a confounding variable.

Subjects are randomised to balance confounding variables between the two populations under study. It’s possible to control for the effect of known confounding variables with statistical analysis, but the advantage of randomisation is that unknown confounding variables should be equally balanced between the two groups.

Randomisation aims to balance the unknown and known confounders.