The change I made should have improved the waiting time. Why did it make it worse?
Simulation includes the generation of passengers, which includes a random element. So, small changes can make counter-intuitive differences in simulation results if your sample size is not big enough. In Elevate you can increase the sample size by increasing the number of simulations run for each configuration in Analysis data.
For example, if you improve door times slightly, you may find that the average waiting time increases when you expect it to decrease. Or you might reduce the lift speed and find that the average waiting time gets less. In some instances, an individual passenger who would have caught the elevator misses it. Of course, if you observe enough passengers, the better performing elevator configuration will perform better. But some “unlucky passengers” can make the average result trend counter-intuitively.
To confirm this, just increase the number of simulations run for each configuration. Sometimes you may have to run as many as 100 simulations to prove the point. You should not need to run this many simulations for design purposes, as small differences in average waiting time (e.g. fractions of seconds) would not be noticeably in the real world.