A detailed study of the standard rounding rule for multiplication and division is presented including its derivation from a basic assumption. Through Monte-Carlo simulations, it is shown that this rule predicts the minimum number of significant digits needed to preserve precision only 46.4% of the time and leads to a loss in precision 53.5% of the time. An alternate rule is studied and is found to be significantly more accurate than the standard rule and completely safe for data, never leading to a loss in precision. It is suggested that this alternate rule be adopted as the new standard.