How this works
Mean, median, and mode are the three classic measures of central tendency — single numbers that summarise where a dataset "sits" on a number line. They answer the same question ("what's typical?") in three different ways, and they often disagree, which is the whole point of having three. The mean is the arithmetic average: sum the values, divide by the count. The median is the middle value when sorted: half the data is above, half below. The mode is the most frequently occurring value (a dataset can have multiple modes if several values tie for the most frequent, or no mode if every value is unique).
Which one to use depends on your data's shape and what you're trying to communicate. The mean is the most informative when the data is roughly symmetric and not heavily skewed — for normally distributed data the mean, median, and mode all coincide. But the mean is sensitive to outliers: one billionaire in a room of 99 ordinary people pulls the mean income way up while leaving the median essentially unchanged. So when reporting "typical" income, "typical" house price, or "typical" anything where the distribution has a long tail (income, wealth, response times, file sizes), use the median — it's a better summary of what most data points look like. The mean is useful when you need to do further math (multiplying, summing across groups, statistical tests like t-tests assume mean-centred data), but it can mislead when reporting a single typical value to a non-technical audience.
The mode is most useful for categorical data ("most popular colour", "most common occupation", "most-clicked button") and for discrete data with natural clustering. For continuous data (heights, temperatures, prices), the mode is often meaningless because no two values are exactly equal — every height is unique to the millimetre. In that case, you'd bin the data into ranges and report the modal bin instead. A practical heuristic: if your dataset is symmetric and you need a single representative number, use the mean. If the data is skewed or has outliers, use the median. If you're describing a distribution's peak or talking about categories, use the mode. Reporting all three and letting the audience compare is often the most honest approach — it surfaces skew when the mean and median disagree.
The formula
xᵢ are the individual data points. n is the count. For median, the data must be sorted first; for an even count of values, the median is the average of the two middle values. Mode requires no sorting and may return multiple values (bimodal, multimodal) or none (when all values are unique).
Example calculation
- Dataset: 7, 3, 9, 3, 5, 8, 3. n = 7.
- Mean = (7+3+9+3+5+8+3) / 7 = 38 / 7 ≈ 5.43.
- Sort: 3, 3, 3, 5, 7, 8, 9. n is odd (7), so median = the (7+1)/2 = 4th value = 5.
- Mode: 3 appears three times, more than any other value, so mode = 3. The mean (5.43), median (5), and mode (3) are all different — typical for a small skewed dataset; reporting all three gives a fuller picture than any single one.
Frequently asked questions
Why are the mean and median often different?
Because the mean is sensitive to outliers and skew, while the median ignores them. The mean uses every value's actual magnitude in the calculation, so a single very large or very small value pulls it noticeably; the median only uses the position of values, so extreme outliers contribute exactly the same as any other point above or below the middle. When mean > median, the data is right-skewed (a long tail of high values pulls the mean up — typical of income, wealth, response times, file sizes). When mean < median, the data is left-skewed (less common, but seen in things like age at retirement or test scores when there's a floor or ceiling effect). The size of the gap between mean and median is itself a quick measure of skewness.
What if my dataset has more than one mode?
Report all of them. A bimodal dataset has two values tied for most frequent (e.g. 3, 3, 5, 5, 7 — both 3 and 5 occur twice, no other value occurs more than once); a multimodal dataset has three or more. The calculator above lists every tied value rather than picking arbitrarily. Bimodality often signals that two distinct subpopulations are being analysed together — for example, the heights of a mixed group of adult men and women is bimodal because each sex has its own typical range. When you see multiple modes, ask whether your dataset is actually mixed and would be better analysed in subgroups. If every value is unique (common with continuous measurements), there's no mode at all — that's the right answer in that case, and you should describe the distribution by mean and median instead.
When should I report all three vs just one?
Always compute all three; report based on audience and purpose. For a technical audience or a thorough analysis, reporting all three with their gaps tells a richer story about the distribution's shape. For a non-technical audience needing one number: median for skewed/income-like data, mean for symmetric/measurement data, mode only when the data is categorical or has natural clustering points. In journalism and policy reporting, the median is usually the right default — it's harder to manipulate with outliers and represents the "middle person" honestly. In scientific reporting, the mean is conventional alongside its standard deviation. In product/UX analytics, all three help: mean for understanding overall usage, median for typical-user behaviour, mode for identifying the most common pattern (e.g. "most users send 3 messages per session" — that's the mode of session message counts).