Abstract
The aim of this essay is to describe the mechanism, which operates in the estimation of the work packages length in a project, composed of independent activities performed simultaneously. Stadistics operations are performed to demonstrate that estimations are based on the average lenght of each activity, without taking in account their variations; they are susceptible to be wrong and underestimate their final duration. Likewise, it demonstrates the risk of delay in this type of work packages, is higher than the risk of each individual work. A simulation is done and it proves the exposed ideas.