The mean and standard deviation tell you where a data set is centered and how spread out it is, but they say nothing about its shape. Skewness and kurtosis fill that gap — skewness measures asymmetry, and kurtosis measures tail weight and peakedness — giving you a fuller picture of what a distribution actually looks like.
What skewness tells you
Skewness describes which side of the distribution has the longer tail. A positive skewness means the tail stretches to the right — a handful of unusually high values pull the average up, which is why household income, home prices, and wait times are all typically right-skewed. A negative skewness means the tail stretches to the left, which is less common but shows up in things like age at retirement or exam scores capped near a maximum. A skewness near zero means the distribution is roughly symmetric, like heights or measurement errors clustered around a true value.
What kurtosis tells you
Kurtosis describes tail weight — how likely extreme values are — and how sharply the distribution peaks near its center. This calculator reports excess kurtosis, which is raw kurtosis minus 3, so a normal distribution always reads as 0. A positive excess kurtosis (leptokurtic) means more of the data sits in a sharp central peak and more sits in the extreme tails, with less in between — financial returns are a classic example, where calm periods are common but so are rare, severe crashes. A negative excess kurtosis (platykurtic) means the data is more evenly spread out, with a flatter peak and thinner tails, like a uniform distribution.
Why population and sample formulas differ
Just like variance, skewness and kurtosis have separate population and sample formulas. When you only have a sample, the raw (population) formulas are slightly biased estimates of the true population shape, so statisticians apply correction factors — the adjusted Fisher-Pearson coefficient for skewness, and a similar bias-corrected formula for kurtosis — that produce more accurate estimates from limited data. The difference between the two formulas shrinks as the sample size grows; with only a handful of points, the choice of estimator can noticeably change the answer.