Give 100 people with headaches a pill that contains only sugar but looks like medicine, and tell them it is a pain reliever. Many of them will report less pain an hour later. Their relief is not faked: expectation can change how much pain people feel, how alert they are and how hard they try. The response comes from believing in the treatment, not from anything in the pill, and that is the placebo effect.
This creates a problem for experiments. If a drug group improves and a no-pill group does not, part of the gap could be belief. The fix is to give the control group a placebo that looks, tastes and is scheduled exactly like the real treatment. Now both groups expect to get better, and the drug's real effect is the extra improvement beyond the placebo group. In a headache trial where 70% get relief on the drug, 45% on the placebo and 20% with no pill at all, the drug adds 25 points beyond belief, and belief adds 25 points beyond nothing.
Not all improvement in a placebo group is a placebo effect. Colds clear up, pain flares settle, and people often join a study when their symptoms are at their worst, so many would improve anyway. Only a group that gets nothing separates natural recovery from the effect of belief. The placebo effect also matters most for outcomes people feel or control, like pain and energy, and much less for outcomes such as whether a bacterial infection clears.
Think of it like this. The placebo effect is like a runner who believes her new shoes are faster: she may truly run faster, from confidence and effort, even if the shoes are no different. To test the shoes fairly, give other runners identical-looking ordinary shoes, tell them the same story, and see whether the new ones still win.
Core idea. Believing you are treated can change real outcomes, especially felt or reported ones, so a fair experiment gives the control group a look-alike placebo and counts as the treatment's effect only what it adds beyond the placebo group.