This paper introduces a basic methodology which, based on simulations, determines the minimum sample size necessary to use the normal approximation when estimating means from population with bounded symmetric distributions.