A controlled experiment isolates the effect of one variable by comparing a group that receives a treatment against a control group that is identical in every other respect.
The idea
Without a control group, you cannot tell whether an observed change was caused by your intervention or by something else entirely — the passage of time, a seasonal trend, or regression to the mean. A controlled experiment splits subjects into two (or more) groups, applies the treatment to only one, and holds everything else as constant as possible so any difference in outcome can be attributed to the treatment itself.
When to use it
- Testing whether a change (a drug, a feature, a policy) actually causes an effect
- Separating causation from mere correlation
- Any situation where you’re tempted to judge an intervention by before/after numbers alone
How to apply it
- Define a single variable to test.
- Randomly assign subjects to treatment and control groups so the groups start out comparable.
- Apply the treatment to one group only; keep conditions identical otherwise.
- Measure the outcome in both groups and compare the difference.
Watch out for
- Confounding variables that differ between groups besides the one you intended to test
- Small sample sizes that make the result unreliable
- Selection bias in how subjects end up in each group
Related models
- Controlled Placebo — adds a placebo arm to control for the act of receiving treatment itself.
- Blinded Randomized Design — strengthens a controlled experiment by removing expectation bias.
- Scientific Method — the controlled experiment is the core tool of hypothesis testing within this broader process.
Sources
Ronald A. Fisher, “The Design of Experiments” (1935), the foundational text formalizing randomization and control groups in experimental design.