A usual unbiased estimator for a population total, mean or a proportion based on randomized response data from a probability sample has a variance as a 'sum' of a term containing the values of a variable of interest and another term involving the variances of unbiased estimators of the variate values obtained by randomized responses. The first term may be controlled by an appropriately chosen sampling design and a suitably specified sample-size. But it is not easy to get the second term suitably and naturally controlled. Thus it remains a problem to suitably fix a sample-size to control the magnitude of this 'sum'. An exercise is presented to implement this task.