In a completely randomised design, all units are randomly assigned to treatment groups without blocking. Each unit has equal probability of receiving any treatment. The design relies on randomisation alone to balance confounding variables across groups.