Skewed numbers: when the median tells a clearer story

Picture a service counter where most visitors are served within a few minutes, but occasionally someone waits an hour. Data like this are skewed, and the mean is often a poor guide to the typical experience. The long wait drags the mean upward, while most visitors waited far less. The median, which is the middle value once the waits are ranked, usually describes what a typical visitor experienced more accurately.

A made-up set shows the effect. Suppose seven visitors wait 4, 5, 5, 6, 7, 8 and 90 minutes. The median is 6 minutes, and the mean is 125 divided by 7, about 17.9 minutes. Remove the 90-minute wait and the median only falls to 5.5 minutes, while the mean drops to about 5.8 minutes. One extreme value moved the mean by about 12 minutes and the median by half a minute.

Spot the long tail

Statisticians describe a distribution by its shape. A distribution is right-skewed, or positively skewed, when its long tail stretches toward high values. Waiting times, household incomes and many house prices behave this way. It is left-skewed when the tail reaches toward low values, as with test scores on an easy exam that most people did well on. In a right-skewed set the mean is usually larger than the median, because the high values add to the total without moving the middle. In a left-skewed set the reverse tends to hold.

You can often spot skew without calculating anything. Sketch the values on a simple dot plot or histogram and see which way the tail runs. Then compare the mean and median. If they differ noticeably, the data are probably not symmetric. A large gap is a signal to report the median, or to report both figures with a note about the tail.

Why the median describes a typical case

The median is resistant to extreme values. It depends on how many values fall below the middle, not on how far the largest value sits above it. If the 90-minute wait were recorded as 900 minutes, the mean would climb even higher, but the median would still be 6. That resistance makes it a fair summary of the typical case. Half the visitors waited less than the median and half waited more, a plain statement a reader can check against the list.

The median is the 50th percentile, the value that half of the cases fall below. Other percentiles follow the same logic, and they are often reported in pairs to describe how the data spread out.

The median has limits too. It ignores how large the values beyond the middle are, so it says nothing about how bad the longest waits were. Two very different data sets can share a median. That is why the median is best read alongside something that describes the spread.

When the mean still matters

The mean is still the right number when totals matter. If you want to know how much time the counter consumed in total, add the individual waits or multiply the mean by the number of visitors. Budgets, tax revenue and total costs are built on sums, and the mean connects those sums to per-person figures. It also reflects the extreme values, which is exactly what a total needs. Dropping the 90-minute wait from the total would understate the time the counter actually spent serving people.

The mean is also the better choice when the distribution is roughly symmetric. In that case the mean and median sit close together, and the mean uses every value. Choosing the mean for symmetric data and the median for skewed data is a simple rule that keeps the summary honest.

Read a report that gives only one average

Reports often present a single average without saying which one it is, or without any hint of skew. When you see one figure, look for a range or percentiles. A common and useful pair is the median together with the quartiles, the points below which one quarter and three quarters of the cases fall. Together they describe the middle half of the data without letting a few extreme cases dominate the picture.

Planning shows why this matters. A team that sets expectations from the mean wait may promise a time that most visitors beat, yet some visitors still wait far longer than the promise suggests. A team that sets them from the median will describe the typical visitor well and say little about the rare long wait. A report that gives the median together with a high percentile, such as the wait that only one visitor in ten exceeds, lets readers judge both the typical case and the tail.

Choosing between them in practice

Start by asking whether the quantity is likely to have a long tail, as waits, incomes and prices usually do. Then ask whether the question is about the typical case or the total. A question about what a typical person experienced points toward the median, while a question about total cost or total demand points toward the mean. When a single figure is all you have, the answers to those two questions usually tell you which number to trust, and which one to ask the publisher for.