Distribution

A distribution shows how a metric's values are spread across users or events — how many have low values, how many high, and the overall shape of that spread.

A distribution is the antidote to 'average-itis': instead of compressing thousands of users into one number, it shows behavior's true shape. Most web metrics have a skewed, long-tailed distribution, not a symmetric bell; the mass of users sits at one end of the range and a small long tail at the other. Seeing that shape changes decisions: a distribution with two peaks (very fast users and very slow users, say) is completely hidden by a single average and lulls you into the illusion of an 'average user'. This is why, before choosing a mean or median, you should first look at the distribution. It is usually viewed with a histogram or box plot, where peaks, gaps and tails become visible.