Agreed. I would also argue that the audience will significantly effect the 'width' of the distribution. For example, lets imagine we're building a model that w

Agreed. I would also argue that the audience will significantly effect the 'width' of the distribution. For example, lets imagine we're building a model that will be used to predict the hosting costs for a website for tweens. I think it would be fair to assume that most tweens will be in bed by 9PM, and can only start accessing your site after school from 4PM. Note im in the UK, school finishes around 4ish... Given that limited visiting window, we might assume that the normal distribution would have a mean around 6:30PM and a standard deviation of around 2, to 2.5 hours. In this way 68% of our traffic would occur in the 4PM while 9PM - or within one standard deviation from the mean.