• Control other variables not of primary interest.
• Use randomization to avoid selection bias and make the groups as similar as possible.
• Replication ensures similar groups and increases the chances of detecting true differences among treatments.