Randomness, seeds, and replications
Understand why runs with random inputs differ, what a random seed controls, and why one run is rarely enough.
A deterministic model gives the same result from the same starting conditions. A stochastic model contains random variation, so repeated runs can differ even when every setting is unchanged. In these guides, stochastic simply means that the model uses randomness.
Arrivals, service times, failures, routes, and human behavior are common sources of variation.
Save the model, scenario, run settings, and seed together. Reusing them lets you repeat the same run; changing the seed gives you another possible outcome from the same setup.
A single run can be unusually lucky or unlucky. Replications repeat the same experiment with different random seeds, revealing how much the result varies from run to run.
Use enough replications to reach the precision the decision requires, and report the spread or an interval alongside the average.
The settings stay the same; only the random stream changes between runs.
Choose a distribution that matches what is known about a quantity's range, typical values, and tails. Use observed data when it is representative; otherwise state the assumption behind the chosen shape.