Abstract
An alternative method of estimating the variance of a PERT activity time is proposed given that one of the initial estimates is the mean. The beta-distribution of the activity time is assumed to be tangential to the x-axis at each end. Restrictions in the current method are relaxed, and the inconsistency of the implicit assumption of a fixed coefficient of skew removed. The skew of the distribution is determined by the three estimates of the activity time. No difficult computations are involved.