“The average is 62” sounds like a fact, and it is closer to a choice: three different numbers all claim the title of a dataset’s center, and picking the wrong one is the polite way statistics mislead. Mean, median and mode answer different questions, disagree exactly when the data is interesting, and take two minutes to keep straight forever. This guide is those two minutes plus the judgment calls, with our free statistics calculator producing all three from any pasted list.
In this guide
Three centers, one dataset
Take seven monthly salaries, in thousands: 30, 35, 35, 40, 45, 50, 200. The mean sums and divides: 435 ÷ 7 ≈ 62.14. The median sorts and takes the middle value: 40, three values below it, three above. The mode is the most frequent value: 35, the only repeat. One dataset, three honest centers, spread from 35 to 62, and the spread itself is the diagnosis: when the three agree, the data is symmetric and boring; when they split this far, something in the data, here the 200, is pulling.
The outlier test: where they split
The split has a direction worth reading. The mean is the only one of the three that every value touches, so a single extreme drags it hard: remove the 200 and the mean collapses from 62.14 to 39.17, while the median barely shifts and the mode does not move at all. That fragility is not a flaw but a feature with a use: the mean detects that an extreme exists, and the median resists it. The classic line explains every salary negotiation graph: when one founder walks into a room of seven people, the mean income jumps and the median does not, and “average salary” in a job ad means whichever number serves the writer. Reading both, and asking why they differ, is the entire skill.
Which one to use, by situation
| Data | Use | Because |
|---|---|---|
| incomes, house prices, anything skewed | median | resists the long tail that drags means |
| physical measurements, repeated readings | mean | uses all information; errors cancel |
| shoe sizes to stock, menu items, votes | mode | “most common” is literally the question |
| ratings on a 1-5 scale | median + distribution | a 3.0 mean can be all 3s or half 1s, half 5s |
The ratings row generalizes into the habit that separates careful readers from headline readers: a center without its distribution hides bimodal stories, the loved-or-hated product looking identical to the uniformly mediocre one. Whole-number datasets get their counts, gaps and repeats profiled by the integer analyzer, which is exactly the “show me the distribution” companion to the headline number.
The center is half the story: spread
Two cities can share a mean temperature of 15°C, one drifting between 10 and 20 all year and one swinging from minus 10 to 40. The numbers that finish the story: the range (max minus min, crude but instant), and the standard deviation, the typical distance of a value from the mean, small when data hugs the center and large when it sprawls. Standard deviation earns its place with one practical reading: for roughly bell-shaped data, about two-thirds of values sit within one standard deviation of the mean and about 95 percent within two, which turns an abstract number into “where do most cases actually live”. The calculator reports both alongside the three centers, which is the full kit for a first honest look at any list of numbers.
Weighted means: when items are not equal
The plain mean assumes every value deserves an equal vote, and plenty of real averages do not work that way: a course grade where the final exam counts triple, a portfolio return where positions differ in size, a customer satisfaction score across stores of different traffic. The weighted mean multiplies each value by its weight, sums, and divides by the total weight, so two exam scores of 90 and 60 weighted 3-to-1 give (90 × 3 + 60 × 1) ÷ 4 = 82.5, not the naive 75. Forgetting the weights is among the most common spreadsheet errors in business reporting, and the symptom is always the same: the “average” of the subgroup averages disagreeing with the true total. More of the toolbox lives in the calculators hub.
Frequently asked questions
What is the median of an even-sized list?
The mean of the two middle values after sorting: for 10, 20, 30, 40 it is 25. Half the data still sits on each side, which is the property the median exists to provide.
Can a dataset have two modes, or none?
Both happen: two values tied for most frequent make the data bimodal, often a sign two different populations were mixed in one list, and a list with no repeats has no mode at all. The mode’s fragility with continuous data is why it shines mostly for categories and counts.
Why do news reports almost always use the mean?
Habit, and sometimes convenience: the mean is easy to compute from totals without sorting anything. For skewed quantities like income and home prices the median is the more honest center, which is why careful statistics agencies publish both and label them.
Is “average” the same as “mean”?
In careful usage, average is the umbrella and mean is one member, but everyday speech uses average to mean the mean. When a number matters, ask which center was computed; when you publish one, say which you used.