Standard Deviation

Standard deviation says how far the data typically sits from the mean; a number that summarizes a set's spread in the data's own unit.

The mean says where the center is; standard deviation says how far the data spreads around it — without the second, the mean can deceive. Two sets with the same mean but different standard deviations are two separate worlds: one uniform and predictable, the other volatile and risky. A large standard deviation means unstable behavior, and that is what matters in a decision: a channel with a good mean conversion but a high standard deviation is unreliable. But on very skewed distributions (common on the web), standard deviation itself becomes misleading, because like the mean it is sensitive to the tail; there, percentiles are the better tool. Standard deviation is the square root of variance and, unlike variance, is in the data's own unit, which is why it is easier to interpret.