You have a list of numbers and you want to summarize them with a single value. You were taught three ways to do this: mean (average), median (middle value), and mode (most frequent). You probably memorized the definitions in school and haven’t thought about them since.
But these three measures can tell dramatically different stories about the same data. Choosing the wrong one doesn’t just give you a slightly off answer. It can give you a completely misleading one. And people with something to sell or prove choose strategically.
The Mean: Add Them Up, Divide by Count
The arithmetic mean is what people usually call “the average.” Add all the values, divide by the number of values. It’s intuitive, familiar, and the most commonly used summary statistic.
It’s also the most easily distorted.
Example: Five employees at a small company earn $40,000, $45,000, $48,000, $52,000, and $315,000. The mean salary is $100,000.
Does $100,000 represent this group? Not really. Four of the five people earn less than half the “average.” The CEO’s salary dragged the mean up to a number that describes nobody. This is the fundamental weakness of the mean: it’s sensitive to extreme values (outliers). A single very high or very low number can shift the mean far from where most of the data sits.
This is why “average income” figures can be so misleading. In income data, the distribution is heavily skewed, with a relatively small number of very high earners pulling the mean well above what a typical person earns.
When the Mean Works
The mean works well when data is roughly symmetrically distributed and doesn’t have extreme outliers. Heights of adults, test scores in a large class, temperatures over a month: these tend to cluster around a central value without extreme skew, and the mean gives a reasonable summary.
It also works when you need to calculate totals. If you know the mean and the count, you can recover the total (mean × count = sum). No other summary statistic gives you this.
The Median: The Middle Value
The median is the value that splits the data in half: 50% of values are below it, 50% above. If you have an odd number of values, it’s the middle one. If even, it’s typically the average of the two middle values.
Same example: Salaries of $40,000, $45,000, $48,000, $52,000, $315,000. The median is $48,000.
This is much more representative. Most people in this group earn something close to $48,000. The CEO’s salary, no matter how extreme, doesn’t affect the median at all. It could be $315,000 or $3 million and the median would still be $48,000.
This is why economists typically report median household income rather than mean household income. The US median household income is around $75,000. The mean household income is over $100,000. The median tells you what a household in the middle of the distribution earns. The mean tells you a number inflated by the top end.
When the Median Falls Short
The median ignores how far the values are from the center. Consider two datasets:
- Dataset A: 10, 20, 30, 40, 50
- Dataset B: 10, 29, 30, 31, 50
Both have a median of 30. But Dataset A is spread out evenly, while Dataset B is clustered tightly around 30 with two outliers. The median doesn’t distinguish between these; you need a measure of spread (like standard deviation or interquartile range) to see the difference.
The median is also less useful for small samples. With only three or four data points, the median can be one of the actual values and may not represent the “center” in any meaningful way.
The Mode: The Most Common Value
The mode is the value that appears most frequently. Unlike the mean and median, it can be used with non-numeric data (the mode of a list of colors is the most common color).
Example: Shoe sizes sold at a store today: 8, 9, 9, 9, 10, 10, 11, 12, 13. The mode is 9.
For the store deciding what to stock, the mode is actually the most useful number here. The mean shoe size (10.1) and median (10) are fine summaries, but neither directly answers “what size do we sell the most of?”
When the Mode Misleads
The mode can be unstable and arbitrary. Add one more size-10 sale to the list above, and suddenly the mode is either 9 and 10 (bimodal) or whichever sold one more. Remove a single size-9 sale and the mode might change entirely.
For continuous data (like height measured to the decimal), the mode is often meaningless because no two values are exactly the same. It really only works for categorical data or data with repeated values.
How These Get Used to Mislead
Once you understand the differences, you start noticing when people pick the statistic that tells the story they want.
“Our employees earn an average of $95,000”: a company might report the mean, inflated by executive compensation, to look like they pay well. The median might be $62,000.
“The average home price in this neighborhood is $380,000”: a real estate agent might choose mean or median depending on which number is more attractive. In a neighborhood with a few mansions, the mean will be higher. In a neighborhood with a few distressed properties, the median will be higher.
“Most customers rate us 5 stars”: the mode is 5, but the mean might be 3.2 because there’s a cluster of 1-star reviews. “Most common” and “typical” are not the same thing.
This isn’t necessarily dishonest. Sometimes the mean or mode really is the right number for the context. But when someone reports a summary statistic without telling you which one, be skeptical.
Which One Should You Use?
Use the mean when your data is roughly symmetric, doesn’t have extreme outliers, and you might need to calculate totals.
Use the median when your data is skewed (income, home prices, response times) or has outliers that would distort the mean.
Use the mode when your data is categorical (colors, sizes, choices) or you need to know the most common outcome.
Report more than one when the data is complex or you want to give a complete picture. Reporting both mean and median, for instance, immediately tells the reader whether the data is skewed (if they’re far apart, it is).
And always pair your central tendency measure with some indicator of spread. At minimum, the range; ideally the standard deviation or interquartile range. A single summary number without any sense of spread doesn’t tell you nearly enough.
Run Your Own Numbers
Try the Statistics Calculator on ToolzHQ to compute mean, median, mode, standard deviation, and more for any dataset. You can see all three measures side by side and judge which one best represents your data.