STUB When you measure independent timings that should happen at a constant rate/time, use a Cumulative Distribution Function plot to establish if your timings form an exponential curve. If they do not, you have not yet properly separated out all confounding factors (aka., you still have an *inhomogeneous* exponential distribution). Once that is done, and you can be sure every measurement is i.i.d., you now can apply both the mean and the median to characterize the time. If the process is a homogeneous exponential, as show by the CDF plot, the two statistics are expected to be close.