Sample size and running multiple simulations

The duration of a simulation is defined by its start and end times. During this period, passengers are generated following the passenger demand template. A simulation concludes once all generated passengers have reached their destinations.

Multiple simulations can be run using different random number seeds. Based on the demand template, these simulations generate the same total number of passengers but with slight variations in their arrival times. This approach simulates the same building experiencing passenger traffic on different days of the week.

The duration and number of simulations are significant because they determine the total number of passengers generated—and, therefore, the statistical robustness of the results. For example, a 5-minute simulation with a constant demand rate of 12% and a population of 100 people would generate:

12% of 100 = 12 passengers.

A 60-minute simulation would therefore generate:
(60 / 5) × 12 = 12 × 12 = 144 passengers.

If ten 60-minute simulations are run, the total number of passengers becomes:
10 × 144 = 1,440 passengers.

If the first and last 5-minute intervals are excluded from the analysis to remove start and end effects, only the middle 50 minutes of each simulation are considered. This reduces the number of analysed passengers to:
(50 / 60) × 1,440 = 1,200 passengers.

If the number of passengers analysed is too low, the results may not be statistically significant and can sometimes appear counterintuitive. For example, increasing the lift speed might result in a longer average waiting time due to random variation.

By default, an ISO simulation requires a single 2-hour passenger demand profile, while the CIBSE templates require ten simulations. This is usually sufficient. However, in cases of low population or low demand, a greater number of simulations may be necessary.

The number of simulations required can be estimated based on experience or determined through trial. In practice, it is rare for 50 simulations to be insufficient, even for buildings with low population and low demand.