This paper presents a pragmatic approach to the study of the influence of tasks time variability in the performance of linear and parallel assembly lines. Discrete event simulation is used to assess the performance of several configurations for two assembly line balancing problems, and for four different levels of time variability. From the simulation studies, a set of equations were developed allowing the prediction of the assembly line performance, measured by the cycle time for common ALBPs, with considerable accuracy. These equations show the interactions between the sum of average task times, the balancing difficulty of the problem and the tasks time variability, as well as the number of workstations and their level of parallelism. With these equations it's possible to reduce the search for the best candidate solutions (number of workstations and level of parallelism) for a given target cycle time, considering the time variability level of the problem. With this approach, the assembly line designer is able to select beforehand the possible feasible solutions in a more practical way when using simulation software.