“The average is $75,000”, but is it? The mean vs median distinction decides whether an average tells the truth or quietly lies, and mixing them up is one of the most common ways statistics mislead. This guide explains both in plain English, works through real examples, and shows you exactly when each one misleads. Browse more math and stats help in our tools collection.
The mean is what most people picture when they hear “average”: add up all the values and divide by how many there are. The median is the middle value when you line every value up in order, half the values sit above it, half below.
Take a small, tidy dataset, the ages at a dinner party: 24, 27, 29, 31, 34.
Here they agree. Now watch what happens when one guest brings their 72 year old grandfather: 24, 27, 29, 31, 34, 72.
One outlier dragged the mean up by more than 7 years, while the median barely moved. That single difference, the mean listens to every value, the median listens only to position is the whole story of when each one misleads.
The mean misleads whenever the data is skewed stretched out on one side by extreme values. Income is the classic case. Imagine nine people earning $40,000 and one person earning $1,000,000:
The mean says the “average” person earns $136,000, a number that describes literally nobody in the group. The median’s $40,000 is what a typical person actually earns. This is why economists, journalists, and housing reports quote median income and median house prices: a handful of mansions or billionaires would otherwise warp the picture beyond recognition.
Other places the mean misleads: investment returns quoted without mentioning a single spectacular year, “average” customer ratings dragged down by a few furious 1 star reviews, and average response times where one catastrophic outage dominates the total.
The median is not a saint either, it misleads by ignoring magnitude. Because it only cares about order, it is blind to how extreme the extremes are. Consider two classrooms’ test scores:
Both classes have a median of 70, but they are clearly different classrooms. Class A has a struggling student and a star, while Class B is tightly clustered. The median erases that entire story. Whenever the spread and the extremes matter, risk, quality control, total cost, the median’s blindness is a liability.
The median also misleads with small datasets. With five values, moving one data point can swing the median wildly, and with an even count the median is an invented midpoint (like our 30 above) that no one actually scored. And the median cannot be combined: you cannot average two medians to get the median of the combined group, while means combine cleanly when you know the group sizes.
Ask two questions. First: are there outliers or heavy skew? If yes, incomes, house prices, wait times, lead with the median. Second: do the extremes matter for the decision? If yes, total budgets, risk exposure, capacity planning. You need the mean (and probably the full distribution too).
The honest move, used by good analysts everywhere, is to report both. “The median delivery time is 2 days, the mean is 3.4” tells you in one breath that service is usually fast but sometimes badly delayed, a story neither number tells alone. When a report gives you only one average, ask which one it is and why the other was left out.
Let us put real numbers through both formulas. Five houses sell on the same street in one year, for $210,000, $230,000, $240,000, $250,000, and $900,000. Four are ordinary family homes; the fifth is a renovated property with a large plot that a developer wanted badly. Which single number honestly describes what a house on this street costs?
The mean first. Add the prices: 210,000 + 230,000 + 240,000 + 250,000 + 900,000 = 1,830,000. Divide by the five sales: 1,830,000 ÷ 5 = $366,000. Now the median. The prices are already in order, so the middle value, the third of five, is $240,000. The two answers sit $126,000 apart. The mean is 52.5% higher than the median (126,000 ÷ 240,000 = 0.525), and you can check that gap yourself with a percentage calculator if you want to see how sensitive it is to the mansion’s price.
Which one would mislead a buyer? A buyer budgeting around the “average” of $366,000 would expect a street of uniformly pricey homes and would be confused to find four of the five sales clustered between $210,000 and $250,000. The median’s $240,000 describes the typical transaction almost exactly. Now remove the developer’s purchase and recompute the mean of the remaining four: 930,000 ÷ 4 = $232,500, suddenly very close to the median. That is the cleanest demonstration of the whole guide: one value out of five moved the mean by more than $130,000, and the median would have moved barely at all. If you are doing this kind of checking by hand, a scientific calculator keeps the long additions honest.
The same shape appears in a workplace version. Six colleagues compare salaries: $45,000, $48,000, $50,000, $52,000, $55,000, and the founder at $180,000. The mean is 430,000 ÷ 6 ≈ $71,667; the median is the average of the two middle values, (50,000 + 52,000) ÷ 2 = $51,000. If a recruiter advertises that the team’s “average salary” is about $72,000, every ordinary employee on that team earns far less than the advertised figure, technically true, practically a lie. The median tells the new hire what a typical colleague actually makes.
The table below lines the two measures up across the situations where the choice matters most. Every figure is computed from the small datasets described, so you can verify each one.
| Dataset | Mean | Median | Which to lead with, and why |
|---|---|---|---|
| House prices: 210k, 230k, 240k, 250k, 900k | $366,000 | $240,000 | Median, one luxury sale inflates the mean by 52.5% |
| Team salaries: 45k, 48k, 50k, 52k, 55k, 180k | About $71,667 | $51,000 | Median, the founder’s salary describes nobody else |
| Delivery times in days: 12, 14, 15, 15, 16, 18, 95 | About 26.4 days | 15 days | Both, the gap itself reveals one disastrous delivery |
| Symmetric test scores centred on 70 | 70 | 70 | Either, with no skew, the two agree |
Read the delivery row carefully, because it is the subtle one. The median of 15 days says a typical order arrives in about two weeks, which is true. The mean of about 26.4 days is dragged up by a single 95 day disaster. Neither number is the whole story: a business that only quoted the median would hide its worst failure, and one that only quoted the mean would look slower than it usually is. Reporting both, with the reason for the gap, is what an honest summary looks like.
Try the distinction on one fresh dataset before you trust that it has stuck. A clinic reports that the average wait to see a doctor is 40 minutes. The five waits recorded that morning were 10, 15, 20, 25, and 130 minutes. Check the claim: the mean is (10 + 15 + 20 + 25 + 130) ÷ 5 = 200 ÷ 5 = 40 minutes, so the clinic’s figure is arithmetically honest. The median, though, is the middle value, 20 minutes. Four of the five patients waited 25 minutes or less; the fifth waited over two hours, and that single experience pulled the published average up to double the typical wait.
Now notice that both summaries fail someone. A patient planning around the median’s 20 minutes has no warning that a two hour wait is possible at this clinic. A patient planning around the mean’s 40 minutes over prepares for a wait that most people never experience. The complete description needs the range as well, “most people wait 10 to 25 minutes; occasionally, much longer”, which is why experienced analysts treat mean and median as the beginning of a summary, not the whole of it. Whenever a single average is doing heavy persuasion work in an advertisement, a headline, or a report, run this test in your head: what would the other average say, and what is the spread hiding? The answers take thirty seconds and routinely change what the number means.
The mean is the sum of all values divided by the count; the median is the middle value when data is sorted. The mean uses every value’s magnitude, so outliers pull it; the median uses only position, so outliers barely move it.
Use the median when data is skewed by outliers, incomes, house prices, and wait times are the classic cases. The median describes the “typical” case better because extreme values cannot drag it around.
Yes, in perfectly symmetric data (like a bell curve) they coincide. When mean and median differ noticeably, that gap itself is information: it tells you the data is skewed, and the direction of the gap tells you which way.
Because a few ultra expensive mansions would inflate the mean far above what a typical buyer pays. The median, the price of the middle house sold, reflects the typical market far more honestly.
No. The median ignores how extreme values are, cannot be combined across groups, and is jumpy in small datasets. When totals matter, budgets, risk, capacity, the mean (plus the spread) is the right tool.