Mean Median Mode Calculator

Understanding the Mean Median Mode Calculator

A calculator giving mean, median, and mode returns all three measures of centre at once, which is the honest approach because they answer different questions. Where they diverge, that divergence is itself the most informative thing about the data.

How it actually works

Enter numbers separated by commas. The calculator sorts them, averages for the mean, takes the middle value for the median, and finds the most frequent value for the mode. The sample data gives a mean of 18 and a median of 15.5 with no mode.

The deeper context most people miss

The mean exceeding the median is the signature of a long upper tail, which is why mean income always exceeds median income. Reporting only the mean for such data describes a value that a majority of the population sits below, which is technically correct and practically misleading.

Why no mode is reported here and what that means

The sample values are all distinct, so no value occurs more often than any other and the mode is undefined, which is the normal situation for continuous data. Exact repeats are rare when values are measured with any precision, so the mode is only meaningful for discrete data, categorical data, or continuous data that has been grouped into bins. That last case introduces a problem: the modal bin depends on where the bin boundaries fall, so the same data binned differently can produce a different mode, which makes it a less stable summary than it appears. Kernel density estimation gives a smoother alternative, estimating the underlying density and locating its peak without arbitrary binning, though it introduces a bandwidth choice with the same character. Where the mode genuinely earns its place is with categorical data, where mean and median are undefined entirely: the most common blood type, the most frequent failure cause, the most popular option. It also matters when the question is specifically about what happens most often rather than what happens typically, which are different questions. A distribution with two distinct peaks is bimodal, and that is important information: it usually indicates two subpopulations mixed together, and reporting any single measure of centre for such data describes a value that may be uncommon in both groups.

A worked example: when the mean sits where nobody is

The sample data gives a mean of 18 and a median of 15.5, a modest gap indicating mild right skew. In genuinely skewed data the gap is large and consequential. For a bimodal distribution the problem is worse still: a dataset of values clustered around 10 and around 90 has a mean near 50, a value that essentially no observation is close to, and the mean describes a point in the empty middle rather than anything real. This is why plotting the data matters more than any summary, and why reporting a mean without indicating the shape can actively mislead. The classic real-world illustration involves the well-known observation that the average person has slightly fewer than two legs, since a small number of people have fewer and essentially nobody has more, so almost everyone is above average. It is a joke that makes a real point about skewed distributions with a hard boundary. The same structure appears in wealth, in company sizes, in city populations, and in the duration of support calls, all of which have long upper tails, and in each case the median describes typical experience while the mean describes the total divided by the count, which is a different and sometimes more relevant quantity.

Deciding which measure to report

Use the median for skewed data and for questions about typical experience, since it is unaffected by how extreme the extremes are and describes the midpoint of the population. Use the mean for symmetric data, and whenever totals matter, since mean times count gives the sum and no other measure has that property, which is why budgeting, capacity planning, and any calculation that must aggregate uses the mean regardless of shape. Use the mode for categorical data where the others are undefined, and when the question is genuinely about the most common case. Report more than one when they differ substantially, since the gap communicates the shape and hiding it is a choice about what the reader is allowed to see. Report a range or specific percentiles alongside, since no single number describes a distribution and the 25th and 75th percentiles communicate spread in a way a centre cannot. Plot the data before choosing, which takes seconds and reveals skew, multiple peaks, and outliers that no summary conveys. And say which measure was used, since average is ambiguous and that ambiguity is exploited routinely in reporting where whichever figure supports the argument gets called the average.

Robustness and the breakdown point

The median is robust and the mean is not, and the difference can be made precise. The breakdown point of an estimator is the proportion of the data that can be made arbitrarily extreme before the estimator itself becomes arbitrarily wrong. For the median it is 50%: up to half the data can be corrupted without the median leaving the range of the remaining good values. For the mean it is zero: a single value taken to infinity takes the mean with it. That difference matters whenever data may contain errors, contamination, or genuine extremes, which is most real data. Trimmed means discard a proportion from each end before averaging and sit between the two, with a breakdown point equal to the trimming proportion, and they are used in competition scoring and some economic statistics precisely for that reason. Winsorised means cap rather than discard extremes. The median absolute deviation is the robust counterpart to the standard deviation, with the same 50% breakdown point. The cost of robustness is efficiency: when the data really is well behaved and free of contamination, the mean uses the information more effectively and has lower variance than the median, so robustness is insurance paid for in precision. Choosing between them is therefore a judgement about how much you trust the data.

Variations: other centres and percentile measures

The geometric mean multiplies values and takes the nth root, and is correct for growth rates and multiplicative processes where the arithmetic mean overstates. The harmonic mean is the reciprocal of the mean of reciprocals and is correct for rates over a fixed distance, which is why averaging two speeds arithmetically gives the wrong answer. Weighted versions of all three apply where observations carry different importance. The midrange averages the extremes and is maximally sensitive to outliers. Trimmed and Winsorised means trade robustness for efficiency. Percentiles and quartiles describe position, with the median being the 50th percentile. The interquartile range describes spread robustly. For grouped data, the mean can be estimated from class midpoints with some loss, and the modal class replaces the mode. For circular data including angles and times of day, ordinary averaging fails, since the mean of 350 degrees and 10 degrees is 180 rather than 0, and circular statistics uses vector methods instead, which is a genuine trap when averaging bearings, wind directions, or clock times.

Choosing and reporting a measure of centre

Plot the data before summarising it, which reveals skew, multiple peaks, and outliers that no single number communicates. Use the median for skewed data and questions about typical experience, since it is unaffected by how extreme the extremes are. Use the mean when totals matter, since mean times count gives the sum and no other measure has that property. Use the mode for categorical data, where mean and median are undefined entirely. Report more than one measure when they differ substantially, since the gap itself tells the reader about the shape. Say which measure you used, since average is ambiguous and the ambiguity is exploited routinely. Report spread alongside centre, since the 25th and 75th percentiles communicate more than any single value. Treat a bimodal distribution as two populations rather than averaging across the division. Use the geometric mean for growth rates and the harmonic mean for rates over fixed distances. And use circular statistics for angles and times, since ordinary averaging gives nonsense across the wraparound.

What people get wrong

• Reporting the mean of skewed data as typical, when income, prices, and durations all have long upper tails that pull the mean above what most of the population experiences.
• Reporting any centre for bimodal data, when the value can sit in the empty gap between two clusters and describe an outcome that is uncommon in both.
• Treating the mode as meaningful for continuous data, where exact repeats are rare and the modal bin depends entirely on where the bin boundaries fall.
• Averaging angles or clock times arithmetically, where the mean of 350 and 10 degrees comes out as 180 rather than 0 and circular statistics is required.

Where the math comes from

Mean is the sum divided by the count. Median is the middle value once sorted, or the average of the two middle values for an even count. Mode is the most frequent value, undefined when all values occur equally often. The mean has a breakdown point of zero, meaning one arbitrarily extreme value moves it arbitrarily, while the median's is 50%.

Questions and answers

Related calculators

Hex Color Tools · Pi Day · Equation of Circle · Golden Ratio · Ratio