The average lies
Every time you read "the average American earns X" or "the average home price is Y", someone chose a statistic for you — and that choice changes the story. The mean is easy to compute and easy to abuse. Let's watch it break right in front of you.
1. The billionaire in the bar
Ten ordinary people are in a bar. Their monthly salaries range from $2,000 to $8,000 — the group average is $4,500. Suddenly a billionaire walks in and sits down with them. His monthly income: $50 million. What is the average salary of the 11 people in the bar now?
2. Drag the outlier
Each dot is a person. Move the last person's salary and watch the two markers: the mean chases the extreme, while the median — the middle value — barely budges. The median only cares about position in the order, not magnitude.
Dot plot of salaries with mean and median markers.
3. Same average, different worlds
Two teams with exactly the same average salary. On team A, everyone earns close to it. On team B, some earn very little and some earn a lot. The mean can't tell them apart — standard deviation can: it measures how far values typically sit from the mean.
Two dot plots sharing the same mean with different spreads.
Use the median when your data has extreme values or is skewed — income, home prices, wait times, view counts. Use the mean when data is symmetric and outlier-free — height, temperature, test scores. And always be suspicious of an average published without a standard deviation or median beside it: whoever gives you only the mean may be hiding the shape of the data.
4. How to choose, in practice
The question that settles most cases: does the distribution have a long tail? Income, house prices, waiting times, follower counts and file sizes do — a few very high values and many low ones. There the mean is dragged by the extremes and the median describes the typical case better. Height, weight, temperature and exam scores tend to be symmetric, and the mean serves well.
A ten-second test: compare the mean with the median. If they are close, the distribution is reasonably symmetric and either will do. If the mean is well above the median, there is a right tail — and anyone reporting only the mean is describing a reality most of the sample does not live in.
Where the mean is irreplaceable: when you need the total. A median cannot reconstruct the sum; a mean can — mean × count returns the exact total. That is why payroll, aggregate consumption and revenue are planned from the mean, even when the median better describes the individual.
And the rule that closes the subject: a mean without a median, and either without the spread, is half the information. Two classes with identical averages can be one where everyone scored 7 and one where half scored 4 and half scored 10.
Run your own data
The full calculator returns mean, median, mode, standard deviation and variance for any list of numbers.
Open the Mean & Standard Deviation Calculator →See also: Weighted average · Geometric mean · Lesson: the percentage trap