Median

The median is the middle value of a set of numbers arranged in order from smallest to largest. Put every observation in a line, and the median is the one standing exactly in the center — half the values sit above it, half below. When the count is odd, the median is a single middle number; when it is even, it is the average of the two middle numbers. Because it depends only on rank and position, not on the size of the values, the median ignores how extreme the largest or smallest numbers are. A billionaire moving into a small town barely nudges its median household income, even though the average income leaps. The median answers a plain question: line everyone up, point to the middle, and ask what the typical case looks like.

The median matters because real data is often skewed, and a skewed distribution makes the average misleading. Incomes, home prices, web page load times, customer order values, and session durations all tend to have a long tail — a handful of very large values that stretch the mean upward, away from where most of the data actually sits. In those cases the median describes the center more honestly than the mean does. If half your visitors load a page in under two seconds but a few stragglers take thirty, the median load time reports the two-second reality most users feel, while the mean is dragged toward the slow tail. Reach for the median whenever a few outliers could distort the picture and you want the value that represents the ordinary case.

The mean and the median are both measures of the center, but they are built differently and can tell different stories. The mean adds up every value and divides by the count, so each observation pulls on it in proportion to its size — one gigantic number can haul the mean far from the bulk of the data. The median only cares about order and position, so that same gigantic number moves it barely at all. In a symmetric distribution the two roughly agree. In a skewed one they diverge, and the gap between them is itself a clue: when the mean sits well above the median, the data has a long right tail of high values, as incomes and prices usually do. Knowing which one you are reading, and why they differ, prevents a great deal of misreading.

Choosing between them is a judgment about what you want to represent and how outliers should count. Use the mean when values are roughly symmetric and every observation genuinely should weigh in — total revenue divided by orders, for instance, where the sum matters. Use the median when the distribution is skewed or noisy and you want the typical case protected from extremes, which is why median salary and median home price are the standard reported figures. The median is called robust because a few corrupted or extreme readings barely shift it, while they can wreck a mean. The trade-off is that the median throws away information about magnitude — it does not know or care that the top value is enormous. Report both when you can, because together they reveal a shape either one alone would hide.

Use the median as your default center whenever data is skewed, bounded at zero, or prone to outliers — money, time, and counts almost always qualify. Pair it with percentiles to describe the whole distribution rather than a single point: the 25th, 50th (the median), and 75th percentiles together sketch the spread, and reporting a median load time alongside the 90th or 95th percentile tells you both the typical experience and the painful tail. When you compare groups, comparing medians is often fairer than comparing means, because one group's stray extreme value will not tilt the contest. And always state your sample size and your definition, because a median computed on a handful of points is fragile, and a median of medians is not the median of the pooled data.

The traps are predictable. People report a mean for skewed money or time data and then wonder why the average feels wrong to everyone living the median reality. They treat the median as if it captured magnitude, forgetting it is blind to how large the extremes are — two datasets with the same median can have wildly different spreads. They compute a median on too few observations and read stability into noise, or they average several group medians and call the result a grand median, which it is not. The discipline is to pick the center that matches the question, quote the median with percentiles so the spread is visible, and remember that the median's strength, its indifference to extremes, is also the information it deliberately discards.

Median comes from the Latin medianus, meaning in the middle, and names the central value of ordered data.

Etymology: source.

Search interest for this term over the last five years:

View interest-over-time on Google Trends →

Areas of marketing where median is a core concern: