Randomness is simulated by the use of probabilities for sequence flows and token routing and also by using statistical distributions to reflect variability in process times of activities etc. To make sure results are valid, the simulation needs be run for long enough to yield random behavior without chance (consider the scenario of tossing a coin or rolling a dice multiple times). Provision should be made to compare results from the same scenario, but different run lengths or replications.The required run length to yield usable outcomes depends on the process model structure, amount of variability and the objective; consequently, a single recommended run length cannot be provided. A replication shares the same scenario configuration and runs for the same length of time, but uses an alternative random stream.
Simulation is well known for providing what-if analysis capabilities; a single simulation run can provide valuable insight on the performance of a particular scenario. The simulation of multiple scenarios and the possibility to compare key outcomes, adds further value and support to decision makers.