A measure of center is a single number that represents the middle or typical value of a dataset. Rather than looking at every individual data point, a measure of center summarizes the entire set with one representative value. The three main measures of center are the mean, the median, and the mode, and each works best in different situations.
The Three Measures of Center
Each measure of center captures “the middle” of a dataset in a different way. The one you should use depends on the type of data you have and whether extreme values are present.
Mean
The mean is what most people think of as the average. You calculate it by adding up all the values in your dataset and dividing by the number of values. If five students scored 70, 80, 85, 90, and 95 on a test, the mean is (70 + 80 + 85 + 90 + 95) รท 5 = 84.
The mean uses every value in the dataset, which makes it very informative but also vulnerable to being pulled by extreme numbers. If one student scored 20 instead of 70, the mean drops to 74, even though most students still scored in the 80s and 90s. This sensitivity to outliers is the mean’s biggest limitation.
Median
The median is the middle value when all data points are lined up from smallest to largest. In the set 70, 80, 85, 90, 95, the median is 85 because it sits right in the center. If you have an even number of values, the median is the average of the two middle numbers.
The median is less affected by outliers than the mean. Change that 70 to a 20, and the median is still 85. This is why you almost always see home prices reported as medians rather than averages. A few multimillion-dollar houses in a neighborhood would drag the mean up dramatically, giving a misleading picture of what most homes actually cost.
Mode
The mode is the value that appears most often. In the set 3, 5, 5, 7, 9, the mode is 5. A dataset can have more than one mode (if two values tie for most frequent) or no mode at all (if every value appears the same number of times).
The mode is the only measure of center that works for non-numerical data. If you surveyed 100 people about their eye color, you couldn’t calculate a mean or median, but you could identify brown as the most common response. Whenever your data consists of categories rather than numbers, the mode is your only option.
How Skewed Data Shifts the Measures
When data is perfectly symmetrical, like a bell curve, the mean, median, and mode all land on the same value. Real-world data rarely looks like that. Most datasets are skewed, meaning they have a longer tail on one side.
In a right-skewed distribution (a long tail stretching toward higher values), the mean gets pulled toward that tail and ends up larger than the median. Hospital length-of-stay data is a classic example: most patients stay a few days, but a small number stay for weeks or months. Those long stays drag the mean well above the median, making the mean a poor representation of the typical patient’s experience. Medical researchers analyzing this kind of data often report the median and interquartile range instead, because the mean and standard deviation are highly sensitive to those outliers.
In a left-skewed distribution (a long tail toward lower values), the opposite happens. The mean is pulled below the median. Retirement age data can look like this: most people retire around 65, but some leave the workforce much earlier due to disability or other circumstances.
When to Use Each Measure
Choosing the right measure of center comes down to two questions: what type of data do you have, and are there outliers?
- Use the mean when your data is numerical and roughly symmetrical, without extreme outliers. Test scores in a large class, daily temperatures over a month, or the weights of items in a shipment are all good candidates. The mean incorporates every data point, giving it the most statistical power of the three measures.
- Use the median when your numerical data has outliers or is noticeably skewed. Income data, home prices, hospital charges, and survival times in medical studies all tend to have long tails that distort the mean. In one analysis of hospital data, the mean values for both length of stay and total charges were much larger than the medians, confirming that a handful of extreme cases were pulling the average away from what was typical. The median gave a more honest picture.
- Use the mode when your data is categorical. Favorite colors, blood types, most common diagnoses in a clinic, or preferred brands in a survey are all situations where mean and median don’t apply.
Why the “Average” Can Be Misleading
In everyday language, people use “average” to mean the mean. But in statistics, the word is more flexible, and this gap causes real confusion. When a news headline says the average American household earns a certain amount, you should ask whether that’s a mean or a median, because the answer can differ by tens of thousands of dollars. Income data is right-skewed: a small number of very high earners pull the mean above what a typical household actually makes. The median household income is almost always lower than the mean, and it better reflects what most families experience.
The same issue shows up in health data. If a study reports that patients spent an average of 8 days in the hospital, a few patients with month-long stays could be inflating that number. When the standard deviation (a measure of spread) is larger than the mean itself, that’s a strong signal the mean is being distorted. In those cases, researchers typically switch to reporting the median because it ignores extreme values and focuses on the true midpoint of the data.
Using Multiple Measures Together
You don’t always have to pick just one. Comparing the mean and median side by side tells you something about the shape of your data. If they’re close together, the data is fairly symmetrical. If the mean is noticeably higher than the median, you’re dealing with a right skew. If the mean is lower, the data skews left. This comparison is a quick diagnostic tool that helps you decide which single number best represents the dataset.
The mode adds a different dimension. In a survey where most respondents chose one answer but a few chose something completely different, the mode tells you the most popular response, while the mean or median (if the data is numerical) tells you where the center of gravity lies. Together, these measures give you a more complete picture than any one of them can alone.